Edition 4.

The GNU C Programming Tutorial

Mark Burgess
Faculty of Engineering, Oslo College

Ron Hale-Evans
Copyright c 2002 Free Software Foundation, Inc.
Permission is granted to copy, distribute and/or modify this document under
the terms of the GNU Free Documentation License, Version 1.1 or any later
version published by the Free Software Foundation; there being no Invariant
Section, with the Front-Cover Texts being “A GNU Manual”, and with the
Back-Cover Texts as in (a) below. A copy of the license is included in the
section entitled “GNU Free Documentation License”.
(a) The FSF’s Back-Cover Text is: “You have freedom to copy and modify
this GNU Manual, like GNU software. Copies published by the Free Software
Foundation raise funds for GNU development.”
Preface xi

Preface
This book is a tutorial for the computer programming language C. Unlike BASIC or Pascal,
C was not written as a teaching aid, but as a professional tool. Programmers love C!
Moreover, C is a standard, widely-used language, and a single C program can often be
made to run on many different kinds of computer. As Richard M. Stallman remarks in
GNU Coding Standards, “Using another language is like using a non-standard feature: it
will cause trouble for users.” (See http://www.gnu.org/prep/standards_toc.html.)
Skeptics have said that everything that can go wrong in C, does. True, it can be unfor-
giving, and holds some difficulties that are not obvious at first, but that is because it does
not withhold its powerful capabilities from the beginner. If you have come to C seeking a
powerful language for writing everyday computer programs, you will not be disappointed.
To get the most from this book, you should have some basic computer literacy — you
should be able to run a program, edit a text file, and so on. You should also have access to
a computer running a GNU system such as GNU/Linux. (For more information on GNU
and the philosophy of free software, see http://www.gnu.org/philosophy/.)
The tutorial introduces basic ideas in a logical order and progresses steadily. You do
not need to follow the order of the chapters rigorously, but if you are a beginner to C, it is
recommended that you do. Later, you can return to this book and copy C code from it; the
many examples range from tiny programs that illustrate the use of one simple feature, to
complete applications that fill several pages. Along the way, there are also brief discussions
of the philosophy behind C.
Computer languages have been around so long that some jargon has developed. You
should not ignore this jargon entirely, because it is the language that programmers speak.
Jargon is explained wherever necessary, but kept to a minimum. There is also a glossary at
the back of the book.
The authors of this book hope you will learn everything you need to write simple C
programs from this book. Further, it is released under the GNU Free Documentation
License, so as the computers and robots in the fantasies of Douglas Adams say, “Share and
Enjoy!”

The first edition of this book was written in 1987, then updated and rewritten in 1999. It
was originally published by Dabs Press. After it went out of print, David Atherton of Dabs
and the original author, Mark Burgess, agreed to release the manuscript. At the request of
the Free Software Foundation, the book was further revised by Ron Hale-Evans in 2001 and
2002.
The current edition is written in Texinfo, which is a documentation system using a single
source file to produce both online information and printed output. You can read this tutorial
online with either the Emacs Info reader, the stand-alone Info reader, or a World Wide
Web browser, or you can read it as a printed book.
The advantages of C 1

1 Introduction
What is a high-level language? Why is C unusual?
Any sufficiently complex object has levels of detail; the amount of detail we see depends
on how closely we scrutinize the object. A computer has many levels of detail.
The terms low level and high level are often used to describe these layers of complexity
in computers. The low level is buried in the computer’s microchips and microcircuits. The
low level is the level at which the computer seems most primitive and mechanical, whereas
the high level describes the computer in less detail, and makes it easier to use.
You can see high levels and low levels in the workings of a car. In a car, the nuts, bolts,
and pistons of the low level can be grouped together conceptually to form the higher-level
engine. Without knowing anything about the nuts and bolts, you can treat the engine as a
black box: a simple unit that behaves in predictable ways. At an even higher level (the one
most people use when driving), you can see a car as a group of these black boxes, including
the engine, the steering, the brakes, and so on. At a high level, a computer also becomes a
group of black boxes.
C is a high-level language. The aim of any high-level computer language is to provide
an easy, natural way to give a list of instructions (a computer program) to a computer.
The native language of the computer is a stream of numbers called machine language. As
you might expect, the action resulting from a single machine language instruction is very
primitive, and many thousands of them can be required to do something substantial. A
high-level language provides a set of instructions you can recombine creatively and give
to the imaginary black boxes of the computer. The high-level language software will then
translate these high-level instructions into low-level machine language instructions.

1.1 The advantages of C

C is one of a large number of high-level languages designed for general-purpose program-
ming, in other words, for writing anything from small programs for personal amusement to
complex industrial applications.
C has many advantages:
• Before C, machine-language programmers criticized high-level languages because, with
their black box approach, they shielded the user from the working details of the com-
puter and all its facilities. C, however, was designed to give access to any level of the
computer down to raw machine language, and because of this, it is perhaps the most
flexible high-level language.
• C has features that allow the programmer to organize programs in a clear, easy, logical
way. For example, C allows meaningful names for variables without any loss of effi-
ciency, yet it gives a complete freedom of programming style, including flexible ways
of making decisions, and a set of flexible commands for performing tasks repetitively
(for, while, do).
• C is succinct. It permits the creation of tidy, compact programs. This feature can
be a mixed blessing, however, and the C programmer must balance simplicity and
readability.
2 Chapter 1: Introduction

• C allows commands that are invalid in other languages. This is no defect, but a powerful
freedom which, when used with caution, makes many things possible. It does mean
that there are concealed difficulties in C, but if you write carefully and thoughtfully,
you can create fast, efficient programs.
• With C, you can use every resource your computer offers. C tries to link closely with
the local environment, providing facilities for gaining access to common peripherals
like disk drives and printers. When new peripherals are invented, the GNU community
quickly provides the ability to program them in C as well. In fact, most of the GNU
project is written in C (as are many other operating systems).
For the reasons outlined above, C is the preeminent high-level language. Clearly, no
language can guarantee good programs, but C can provide a framework in which it is easy
to program well.

1.2 Questions for Chapter 1

1. Explain the distinction between high levels and low levels.
2. What is a “black box”?
3. Name a few advantages to programming in the C language.
The compiler 3

2 Using a compiler
How to use a compiler. What can go wrong.
The operating system is the layer of software that drives the hardware of a computer and
provides the user with a comfortable work environment. Operating systems vary, but most
have a shell, or text interface. You use the GNU shell every time you type in a command
that launches an email program or text editor under GNU.
In the following sections of this chapter, we will explore how to create a C program from
the GNU shell, and what might go wrong when you do.

2.1 Basic ideas about C

First a note about a programming language that is different from the C programming
language, the GNU shell. When you enter commands in the GNU shell, they are executed
immediately. Moreover, the shell is a programming language, in that the commands you
type are a program, because you can also create a text file containing many shell commands.
When you run this file, the commands will be executed in sequence.
On the other hand, consider C. While a shell command file can be executed directly, a
C program must be created in two stages:
1. First, the program is written in the form of text files with a text editor such as GNU
Emacs. This form of the program is called the source code. A computer cannot execute
source code directly.
2. Second, the completed source code is processed with a compiler — a program that
generates a new file containing a machine-language translation of the source code.
This file is called an executable file, or executable. The executable file is said to have
been compiled from the source code.
To run the compiled program, you must usually type the name of the executable file
preceded by a period and a slash, as in this example:
./myprogram
The “dot-slash” prefix tells the GNU shell to look in the current directory for the executable.
You usually do not need to type ‘./’ in front of commands for programs that came with
your GNU system, such as emacs, because the computer already knows where to look for
the executables of those programs, which were placed in special directories when your GNU
system was installed.
A C program is made up of, among other components, variables and functions. A variable
is a way to hold some data which may vary, hence the name. For example, a variable might
hold the number 17, and later the number 41. Another variable might hold the word “Sue”.
A function is a segment of text in the source code of a program that tells the computer
what to do. Programming consists, in large part, of writing functions.

2.2 The compiler

When you compile a program, the compiler usually operates in an orderly sequence of phases
called passes. The sequence happens approximately like this:
1. First, the compiler reads the source code, perhaps generating an intermediate code
(such as pseudo-code) that simplifies the source code for subsequent processing.
4 Chapter 2: Using a compiler

2. Next, the compiler converts the intermediate code (if there is any) or the original
source code into an object code file, which contains machine language but is not yet
executable. The compiler builds a separate object file for each source file. These are
only temporary and are deleted by the compiler after compilation.
3. Finally, the compiler runs a linker. The linker merges the newly-created object code
with some standard, built-in object code to produce an executable file that can stand
alone.
GNU environments use a simple command to invoke the C compiler: gcc, which stands for
“GNU Compiler Collection”. (It used to stand for “GNU C Compiler”, but now GCC can
compile many more languages than just C.) Thus, to compile a small program, you will
usually type something like the following command:
gcc file_name
On GNU systems, this results in the creation of an executable program with the default
name ‘a.out’. To tell the compiler you would like the executable program to be called
something else, use the ‘-o’ option for setting the name of the object code:
gcc -o program_name file_name
For example, to create a program called ‘myprog’ from a file called ‘myprog.c’, write
gcc -o myprog myprog.c
To launch the resulting program ‘myprog’ from the same directory, type
./myprog

2.3 File names

GCC uses the following file name conventions:

Source code file program name.c

Object file program name.o

Executable file program name (no ending)

Header file name.h

Library file libname.a or libname.so

The file name endings, or file extensions, identify the contents of files to the compiler. For
example, the ‘.c’ suffix tells the compiler that the file contains C source code, and the other
letters indicate other kinds of files in a similar way.

2.4 Errors
Errors are mistakes that programmers make in their code. There are two main kinds of
errors.
• Compile-time errors are errors caught by the compiler. They can be syntax errors, such
as typing fro instead of for, or they can be errors caused by the incorrect construction
Errors 5

of your program. For example, you might tell the compiler that a certain variable is
an integer, then attempt to give it a non-integer value such as 5.23. (See Section 2.4.2
[Type errors], page 6.)
The compiler lists all compile-time errors at once, with the line number at which each
error occurred in the source code, and a message that explains what went wrong.
For example, suppose that, in your file ‘eg.c’ you write
y = sin (x];
instead of
y = sin (x);
(By the way, this is an example of assignment. With the equals sign (‘=’), you are
assigning the variable y (causing the variable y to contain) the sine of the variable x.
This is somewhat different from the way equals signs work in mathematics. In math,
an equals sign indicates that the numbers and variables on either side of it are already
equal; in C, an equals sign makes things equal. Sometimes it is useful to think of the
equals sign as an abbreviation for the phrase “becomes the value of”.)
Ignore the syntactic details of the statements above for now, except to note that clos-
ing the (x) with a square bracket instead of a parenthesis is an error in C. Upon
compilation, you will see something like this error message:
error
eg.c: In function ‘main’:
eg.c:8: parse error before ‘]’
(If you compile the program within Emacs, you can jump directly to the error. We will
discuss this feature later. See Chapter 23 [Debugging], page 207, for more information.)
A program with compile-time errors will cause the compiler to halt, and will not produce
an executable. However, the compiler will check the syntax up to the last line of your
source code before stopping, and it is common for a single real error, even something
as simple as a missing parenthesis, to result in a huge and confusing list of nonexistent
“errors” from the compiler. This can be shocking and disheartening to novices, but
you’ll get used to it with experience. (We will provide an example later in the book.
See Chapter 23 [Debugging], page 207.)
As a rule, the best way to approach this kind of problem is to look for the first error,
fix that, and then recompile. You will soon come to recognize when subsequent error
messages are due to independent problems and when they are due to a cascade.
• Run-time errors are errors that occur in a compiled and running program, sometimes
long after it has been compiled.
One kind of run-time error happens when you write a running program that does not
do what you intend. For example, you intend to send a letter to all drivers whose
licenses will expire in June, but instead, you send a letter to all drivers whose licenses
will ever expire.
Another kind of run-time error can cause your program to crash, or quit abruptly. For
example, you may tell the computer to examine a part of its memory that doesn’t exist,
or to divide some variable by zero. Fortunately, the GNU environment is extremely
stable, and very little will occur other than an error message in your terminal window
when you crash a program you are writing under GNU.
6 Chapter 2: Using a compiler

If the compilation of a program is successful, then a new executable file is created.

When a programmer wishes to make alterations and corrections to a C program, these
must be made in the source code, using a text editor; after making the changes, the pro-
grammer must recompile the program, or its salient parts.

2.4.1 Typographical errors

The compiler can sometimes fail for very simple reasons, such as typographical errors,
including the misuse of upper- and lower-case characters. The C language is case-sensitive.
Unlike languages such as Pascal and some versions of BASIC, C distinguishes between
upper- and lower-case letters, such as ‘A’ and ‘a’. If a letter is typed in the wrong case in a
critical place in the source code, compilation will fail. This is a potential source of errors
that are difficult to find.

2.4.2 Type errors

C supports a variety of variable types (different kinds of variables for different kinds of
data), such as integer for integer numbers, and float for numbers with fractional parts.
You can even define your own types, such as total for a sum, or surname for someone’s
last name. You can also convert a variable of one type into other types. (This is called type
coercion.) Consequently, the type of a variable is of great importance to the compiler.
C requires us to list the names and types of all variables that will be used in a program,
and provide information about where they are going to be used. This is called declaring
variables. If you fail to declare a variable, or use it as if it were a different type from the
type it is declared to be, for example, by assigning a non-integer value to an integer variable,
you will receive a compile-time error.
See Chapter 5 [Variables and declarations], page 17, for more information on variable
declarations. See Chapter 3 [The form of a C program], page 7, for some simple examples
of variable declarations.

2.5 Questions for Chapter 2

1. What is a compiler?
2. How does one run a C program?
3. How does one usually compile a C program?
4. Are upper and lower case equivalent in C?
5. What the two main kinds of error that can occur in a program?
6. If you had some C source code that you wished to call “accounts”, under what name
would you save it?
7. What would be the name of the executable file for the program in the last question?
8. How would you run this program?
A word about style 7

3 The form of a C program

What goes into a C program? What does one look like?
The basic building block of a C program is the function. Every C program is a collection
of one or more functions. Functions are made of variable declarations and statements, or
complex commands, and are surrounded by curly brackets (‘{’ and ‘}’).
One and only one of the functions in a program must have the name main. This function
is always the starting point of a C program, so the simplest C program is a single function
definition:
main ()
{
}
The parentheses ‘()’ that follow the name of the function must be included. This is how C
distinguishes functions from ordinary variables.
The function main does not need to be at the top of a program, so a C program does not
necessarily start at line 1, but wherever the function called main is located. The function
main cannot be called, or started, by any other function in the program. Only the operating
system can call main; this is how a C program is started.
The next most simple C program is perhaps a program that starts, calls a function that
does nothing, and then ends.
/******************************************************/
/* */
/* Program : do nothing */
/* */
/******************************************************/

main() /* Main program */

{
do_nothing();
}

/******************************************************/

do_nothing() /* Function called */

{
}
(Any text sandwiched between ‘/*’ and ‘*/’ in C code is a comment for other humans to
read. See the section on comments below for more information.)
There are several things to notice about this program.
First, this program consists of two functions, one of which calls the other.
Second, the function do_nothing is called by simply typing the main part of its name
followed by ‘()’ parentheses and a semicolon.
Third, the semicolon is vital; every simple statement in C ends with one. This is a signal
to the compiler that the end of a statement has been reached and that anything that follows
is part of another statement. This signal helps the compiler diagnose errors.
Fourth, the curly bracket characters ‘{’ and ‘}’ outline a block of statements. When
this program meets the closing ‘}’ of the second function’s block, it transfers control back
to ‘main’, where it meets another ‘}’, and the program ends.
8 Chapter 3: The form of a C program

3.1 A word about style

The code examples above are simple, but they illustrate the control flow of a C program,
or the order in which its statements are executed. You should note that these programs are
written in “old-fashioned” C, as the language existed before ANSI Standard C — the version
in which most C programs are now written. The above programs are also missing several
key elements that most C programs have, such as header files and function prototypes.
Finally, they do not show good style; if you wish to submit programs you write to the Free
Software Foundation, you should consult its advice on how best to use the C language.
You may wonder why we chose old-style C for these first few examples, even though
people proverbially learn best what they learn first. We did so because pre-ANSI C is con-
siderably simpler than the present form, and also because as you develop as a C programmer,
you will probably run across some old C code that you will want to read.
You may also wonder why a savvy programmer would want to follow the ANSI Standard,
which was drafted by committee, or even the GNU guidelines. Isn’t programming free
software all about freedom? Yes, but following the ANSI Standard ensures that your code
can be easily compiled on many other computer platforms, and the GNU guidelines ensure
that your code can be read by other programmers. (We will introduce good C style in our
examples soon. Meanwhile, you can examine the GNU guidelines later in the book. See
Chapter 22 [Style], page 203.)

3.2 Comments

Annotating programs.

Comments are a way of inserting remarks and reminders into code without affecting its
behavior. Since comments are only read by other humans, you can put anything you wish
to in a comment, but it is better to be informative than humorous.
The compiler ignores comments, treating them as though they were whitespace (blank
characters, such as spaces, tabs, or carriage returns), and they are consequently ignored.
During compilation, comments are simply stripped out of the code, so programs can contain
any number of comments without losing speed.
Because a comment is treated as whitespace, it can be placed anywhere whitespace is
valid, even in the middle of a statement. (Such a practice can make your code difficult to
read, however.)
Any text sandwiched between ‘/*’ and ‘*/’ in C code is a comment. Here is an example
of a C comment:
/* ...... comment ......*/
Comments do not necessarily terminate at the end of a line, only with the characters
‘*/’. If you forget to close a comment with the characters ‘*/’, the compiler will display an
‘unterminated comment’ error when you try to compile your code.

3.3 Example 1
#include <stdio.h> /* header file */
Questions for Chapter 3 9

main () /* Trivial program */

/* This little line has no effect */

/* This little line has none */
/* This little line went all the way down
to the next line,
And so on...
And so on...
And so on... */

do_little();

printf ("Function ’main’ completing.\n");

}

/**********************************************/

/* A bar like the one above can be used to */

/* separate functions visibly in a program */

do_little ()
{

/* This function does little. */

printf ("Function ’do_little’ completing.\n");

}
Again, this example is old-fashioned C, and in mediocre style. To make it more compliant
with the ANSI Standard and GNU guidelines, we would declare the variable type each
function returns (int for main, which also requires an exit or return statement), and we
would create function prototypes at the beginning of the file. (See Chapter 4 [Functions],
page 11.)

3.4 Questions for Chapter 3

1. What is a block?
2. Does a C program start at the beginning? Where is the beginning?
3. What happens when a program comes to a ‘}’ character? What does this character
signify?
4. What vital piece of punctuation goes at the end of every simple C statement?
5. What happens if a comment is not ended? That is if the programmer types ‘/*’ .. to
start but forgets the ..‘*/’ to close?
10 Chapter 3: The form of a C program
Function examples 11

4 Functions

Solving problems and getting results.

A function is a section of program code that performs a particular task. Making functions
is a way of isolating one section of code from other independent sections. Functions allow a
programmer to separate code by its purpose, and make a section of code reusable — that
is, make it so the section can be called in many different contexts.
Functions should be written in the following form:
type function_name (type parameter1_name, type parameter2_name, ...)

{
variable declarations

statements
...
...
...
}

You may notice when reading the examples in this chapter that this format is somewhat
different from the one we have used so far. This format conforms to the ANSI Standard
and is better C. The other way is old-fashioned C, although GCC will still compile it.
Nevertheless, GCC is not guaranteed to do so in the future, and we will use ANSI Standard
C in this text from now on.
As shown above, a function can have a number of parameters, or pieces of information
from outside, and the function’s body consists of a number of declarations and statements,
enclosed by curly brackets: ‘{...}’.

4.1 Function names

Every function has a name by which it is known to the rest of the program. The name of
a function in C can be anything from a single letter to a long word. The ANSI Standard,
however, only guarantees that C will be able to distinguish the first 31 letters of identifiers, or
function and variable names. (Identifiers are therefore said to have 31 significant characters.)
In some cases, identifiers may have as few as six significant characters, to stay compatible
with older linkers, but this part of the ANSI Standard is becoming obsolete.
A function name must begin with an alphabetic letter or the underscore ‘_’ character,
but the other characters in the name can be chosen from the following groups:
• Any lower-case letter from ‘a’ to ‘z’
• Any upper-case letter from ‘A’ to ‘Z’
• Any digit from ‘0’ to ‘9’
• The underscore character ‘_’

Note that with GCC, you can also use dollar signs (‘$’) in identifiers. This is one of
GCC’s extensions to the C language, and is not part of the ANSI standard. It also may not
be supported under GCC on certain hardware platforms.
12 Chapter 4: Functions

4.2 Function examples

Here is an example of a function that adds two integers and prints the sum with C’s “print
formatted” function named printf, using the characters ‘%d’ to specify integer output.
void add_two_numbers (int a, int b) /* Add a and b */
{
int c;

c = a + b;
printf ("%d\n", c);
}
The variables a and b are parameters passed in from outside the function. The code defines
a, b, and c to be of type int, or integer.
The function above is not much use standing alone. Here is a main function that calls
the add_two_numbers function:
int main()
{
int var1, var2;

var1 = 1;
var2 = 53;

add_two_numbers (var1, var2);

add_two_numbers (1, 2);

exit(0);
}
When these functions are incorporated into a C program, together they print the number
54, then they print the number 3, and then they stop.

4.3 Functions with values

In mathematics, a function takes one or more values and calculates, or returns, another
value. In C, some functions return values and others do not; whether a function you write
does or does not will depend on what you want the function to do. For example, a function
that calculates a value should probably return that value, while a function that merely
prints something out may not need to.
The add_two_numbers function above did not return a value. We will now examine a
function that does.
Here is an example of calling a function that returns a value:
bill = calculate_bill (data1, data2, data3);
When this statement is executed, control is passed to the function calculate_bill, that
function executes, and then it returns control and some value to the original statement.
The value returned is assigned to bill, and the program continues.
In C, returning a value from a function is a simple matter. Consider the function
calculate_bill as it might be written in a program that contains the statement above:
int calculate_bill (int a, int b, int c)
{
int total;
Function prototyping 13

total = a + b + c;
return total;
}
As soon as the return statement is met, calculate_bill stops executing and returns the
value total.
A function that returns a value must have a return statement. Forgetting it can ruin a
program. For instance if calculate_bill had read as follows, then the variable bill would
have had no meaningful value assigned to it, and you might have received a warning from
the compiler as well. (The word void below indicates that the function does not return a
value. In ANSI C, you must place it before the name of any such function.)
void calculate_bill (int a, int b, int c)
{
int total;

total = a + b + c;
}
On the other hand, you do not need to actually use a value when a function returns
one. For example, the C input/output functions printf and scanf return values, but the
values are rarely used. See hundefinedi [files], page hundefinedi, for more information on
these functions.
If we use the first version of the calculate_bill function (the one that contains the
line return total;), the value of the function can simply be discarded. (Of course, the
resulting program is not very useful, since it never displays a value to the user!)
int main()
{
calculate_bill (1, 2, 3);
exit (0);
}

4.4 Function prototyping

Functions do not have to return integer values, as in the above examples, but can return
almost any type of value, including floating point and character values. (See Chapter 5
[Variables and declarations], page 17, for more information on variable types.)
A function must be declared to return a certain variable type (such as an integer),
just as variables must be. (See Chapter 5 [Variables and declarations], page 17, for more
information about variable types.) To write code in good C style, you should declare what
type of value a function returns (and what type of parameters it accepts) in two places:
1. At the beginning of the program, in global scope. (See Chapter 6 [Scope], page 25.)
2. In the definition of the function itself.
Function declarations at the beginning of a program are called prototypes. Here is an
example of a program in which prototypes are used:
#include <stdio.h>

void print_stuff (int foo, int bar);

int calc_value (int bas, int quux);

void print_stuff (int foo, int bar)

14 Chapter 4: Functions

{
int var_to_print;

var_to_print = calc_value (foo, bar);

printf ("var_to_print = %d\n", var_to_print);
}

int calc_value (int bas, int quux)

{
return bas * quux;
}

int main()
{
print_stuff (23, 5);
exit (0);
}
The above program will print the text ‘var_to_print = 115’ and then quit.
Prototypes may seem to be a nuisance, but they overcome a problem intrinsic to com-
pilers, which is that they compile functions as they come upon them. Without function
prototypes, you usually cannot write code that calls a function before the function itself is
defined in the program. If you place prototypes for your functions in a header file, however,
you can call the functions from any source code file that includes the header. This is one
reason C is considered to be such a flexible programming language.
Some compilers avoid the use of prototypes by making a first pass just to see what
functions are there, and a second pass to do the work, but this takes about twice as long.
Programmers already hate the time compilers take, and do not want to use compilers
that make unnecessary passes on their source code, making prototypes a necessity. Also,
prototypes enable the C compiler to do more rigorous error checking, and that saves an
enormous amount of time and grief.

4.5 The exit function

GNU coding standards specify that you should always use exit (or return) within your
main function. (See Chapter 22 [Style], page 203.)
You can use the exit function to terminate a program at any point, no matter how
many function calls have been made. Before it terminates the program, it calls a number
of other functions that perform tidy-up duties such as closing open files.
exit is called with a return code, like this:
exit(0);
In the example above, the return code is 0. Any program that calls your program can read
the return code from your program. The return code is like a return value from another
function that is not main; in fact, most of the time you can use the return command within
your main, instead of exit.
Conventionally, a return code of 0 specifies that your program has ended normally and
all is well. (You can remember this as “zero errors”, although for technical reasons, you
cannot use the number of errors your program found as the return code. See Chapter 22
Questions for Chapter 4 15

[Style], page 203.) A return code other than 0 indicates that some sort of error has occurred.
If your code terminates when it encounters an error, use exit, and specify a non-zero return
code.

4.6 Questions for Chapter 4

1. Write a function that takes two values a and b, then returns the value of a * b (that
is, a times b.)
2. Is there anything wrong with a function that returns no value?
3. What happens if a function returns a value but you do not assign that value to anything?
4. What happens if a variable is assigned the result of a function, but the function does
not return a value?
5. How can you make a program terminate, anywhere in the program?
16 Chapter 4: Functions
Integer variables 17

5 Variables and declarations

Storing data. Discriminating types. Declaring data.
Variable names in C follow the same rules as function names, as far as what characters
they can contain. (See Section 4.1 [Function names], page 11.) Variables work differently
from functions, however. Every variable in C has a data type, or type, that conveys to the
the compiler what sort of data will be stored in it. Functions in C are sometimes said to
have types, but a function’s type is actually the data type of the variable it returns.
In some older computer languages like BASIC, and even some newer ones like Perl, you
can tell what type a variable is because its name begins or ends with a special character.
For example, in many versions of BASIC, all integer variable names end with a percent sign
(‘%’) — for example, ‘YEAR%’. No such convention exists in C. Instead, we declare variables,
or tell the compiler that they are of a certain type, before they are used. This feature of C
has the following advantages (among others):
• It gives a compiler precise information about the amount of memory that will have to
be allotted to a variable when a program is run, and what sort of arithmetic will have
to be used with it (e.g. integer, floating point, or none at all).
• It provides the compiler with a list of the variables so that it can catch errors in the
code, such as assigning a string to an integer variable.
There are a lot of variable types in C. In fact, you can define your own, but there are
some basic types ready for use. We will discuss them in the following sections.

5.1 Integer variables

C has five kinds of integer. An integer is a whole number (a number without a fractional
part). In C, there are a limited number of integers possible; how many depends on the type
of integer. In arithmetic, you can have as large a number as you like, but C integer types
always have a largest (and smallest) possible number.
• char: A single byte, usually one ASCII character. (See the section on the char type
below.)
• short: A short integer (16 bits long on most GNU systems). Also called short int.
Rarely used.
• int: A standard integer (32 bits long on most GNU systems).
• long: A long integer (32 bits long on most GNU systems, the same as int). Also called
long int.
• long long: A long long integer (64 bits long on most GNU systems). Also called long
long int.
64-bit operating systems are now appearing in which long integers are 64 bits long. With
GCC, long integers are normally 32 bits long and long long integers are 64 bits long, but
it varies with the computer hardware and implementation of GCC, so check your system’s
documentation.
These integer types differ in the size of the integer they can hold and the amount of
storage required for them. The sizes of these variables depend on the hardware and operating
18 Chapter 5: Variables and declarations

system of the computer. On a typical 32-bit GNU system, the sizes of the integer types are
as follows.
Type Bits Possible Values

char 8 -127 to 127

unsigned char 8 0 to 255

short 16 -32,767 to 32,767

unsigned short 16 0 to 65,535

int 32 -2,147,483,647 to 2,147,483,647

unsigned int 32 0 to 4,294,967,295

long 32 -2,147,483,647 to 2,147,483,647

unsigned long 32 0 to 4,294,967,295

long long 64 -9,223,372,036,854,775,807 to

9,223,372,036,854,775,807
unsigned long long 64 0 to 18,446,744,073,709,551,615

On some computers, the lowest possible value may be 1 less than shown here; for example,
the smallest possible short may be -32,768 rather than -32,767.
The word unsigned, when placed in front of integer types, means that only positive or
zero values can be used in that variable (i.e. it cannot have a minus sign). The advantage
is that larger numbers can then be stored in the same variable. The ANSI standard also
allows the word signed to be placed before an integer, to indicate the opposite of unsigned.

5.1.1 The char type

char is a special integer type designed for storing single characters. The integer value of
a char corresponds to an ASCII character. For example, a value of 65 corresponds to the
letter ‘A’, 66 corresponds to ‘B’, 67 to ‘C’, and so on.
As in the table above, unsigned char permits values from 0 to 255, and signed char
permits values from -127 (or -128) to 127. The char type is signed by default on some
computers, but unsigned on others. (See Appendix E [Character conversion table], page 233.
See Appendix D [Special characters], page 231.)
char is used only within arrays; variables meant to hold one character should be declared
int. (See Chapter 15 [Strings], page 97, for more information on character arrays. See
Section 5.4.1 [Cast operator demo], page 21, for an example of how to use an integer
variable to hold a character value.)
Declarations 19

5.1.2 Floating point variables

Floating point numbers are numbers with a decimal point. There are different sizes of
floating point numbers in C. The float type can contain large floating point numbers
with a small degree of precision, but the double-precision double type can hold even larger
numbers with a higher degree of precision. (Precision is simply the number of decimal
places to which a number can be computed with accuracy. If a number can be computed
to five decimal places, it is said to have five significant digits.)
All floating point mathematical functions built into C require double or long float
arguments (long float variables are generally the same as double variables on GNU sys-
tems), so it is common to use float only for storage of small floating point numbers, and
to use double everywhere else.
Here are the floating point variable types available in C:
• float: A single-precision floating point number, with at least 6 significant decimal
digits.
• double: A double-precision floating point number. Usually the same as long float
on GNU systems. Has at least 10 significant decimal digits.
• long double: Usually the same as double on GNU systems, but may be a 128-bit
number in some cases.
On a typical 32-bit GNU system, the sizes of the different floating point types are as
follows.
Type Bits Possible values (approx.)

float 32 1e-38 to 1e+38

double 64 2e-308 to 1e+308

long double 64 2e-308 to 1e+308

You may find the figures in the right-hand column confusing. They use a form of short-
hand for large numbers. For example, the number 5e2 means 5 ∗ 102 , or 500. 5e-2 means
5 ∗ 10− 2 (5/100, or 1/20). You can see, therefore, that the float, double, and long double
types can contain some very large and very small numbers indeed. (When you work with
large and small numbers in C, you will use this notation in your code.)

5.2 Declarations
To declare a variable, write the type followed by a list of variables of that type:
type_name variable_name_1, ..., variable_name_n ;
For example:
int last_year, cur_year;
long double earth_mass, mars_mass, venus_mass;
unsigned int num_pets;

long city_pop, state_pop;

20 Chapter 5: Variables and declarations

state_pop = city_pop = 5000000;

short moon_landing = 1969;

float temp1, temp2, temp3;

temp1 = 98.6;
temp2 = 98.7;
temp3 = 98.5;

double bignum, smallnum;

bignum = 2.36e208;
smallnum = 3.2e-300;
Always declare your variables. A compiler will catch a missing declaration every time and
terminate compilation, complaining bitterly. (You will often see a host of error messages,
one for each use of the undeclared variable. See Chapter 23 [Debugging], page 207.)

5.3 Initialization
Assigning a variable its first value is called initializing the variable. When you declare a
variable in C, you can also initialize it at the same time. This is no more efficient in terms
of a running program than doing it in two stages, but sometimes creates tidier and more
compact code. Consider the following:
int initial_year;
float percent_complete;

initial_year = 1969;
percent_complete = 89.5;
The code above is equivalent to the code below, but the code below is more compact.
int initial_year = 1969;
float percent_complete = 89.5;
You can always write declarations and initializers this way, but you may not always want
to. (See Chapter 22 [Style], page 203.)

5.4 The cast operator

An operator is a symbol or string of C characters used as a function. One very valuable
operator in C is the cast operator, which converts one type into another. Its general form
is as follows:
(type ) variable
For example, floating point and integer types can be interconverted:
float exact_length;
int rough_length;

exact_length = 3.37;
rough_length = (int) exact_length;
In the example above, the cast operator rounds the number down when converting it from a
float to an integer, because an integer number cannot represent the fractional part after the
decimal point. Note that C always truncates, or rounds down, a number when converting
it to an integer. For example, both 3.1 and 3.9 are truncated to 3 when C is converting
them to integer values.
The cast operator works the other way around, too:
Storage classes 21

float exact_length;
int rough_length;

rough_length = 12;
exact_length = (float) rough_length;

In converting large integers to floating point numbers, you may lose some precision, since
the float type guarantees only 6 significant digits, and the double type guarantees only
10.
It does not always make sense to convert types. (See Chapter 20 [Data structures],
page 183, for examples of types that do not convert to other types well.)

5.4.1 Cast operator demo

The following is an example of how to use the cast operator in C code. It also shows how
to use an integer variable to store a character value.
/***************************************************/
/* */
/* Demo of Cast operator */
/* */
/***************************************************/

#include <stdio.h>

int main() /* Use int float and int */

{
float my_float;
int my_int;
int my_ch;

my_float = 75.345;
my_int = (int) my_float;
my_ch = (int) my_float;
printf ("Convert from float my_float=%f to my_int=%d and my_ch=%c\n",
my_float, my_int, my_ch);

my_int = 69;
my_float = (float) my_int;
my_ch = my_int;
printf ("Convert from int my_int=%d to my_float=%f and my_ch=%c\n",
my_int, my_float, my_ch);

my_ch = ’*’;
my_int = my_ch;
my_float = (float) my_ch;
printf ("Convert from int my_ch=%c to my_int=%d and my_float=%f\n",
my_ch, my_int, my_float);

exit(0);
}

Here is the sort of output you should expect (floating point values may differ slightly):
Convert from float my_float=75.345001 to my_int=75 and my_ch=K
Convert from int my_int=69 to my_float=69.000000 and my_ch=E
Convert from int my_ch=* to my_int=42 and my_float=42.000000
22 Chapter 5: Variables and declarations

5.5 Storage classes

There are a few variable declaration keywords commonly used in C that do not specify
variable types, but a related concept called storage classes. Two common examples of
storage class specifiers are the keywords extern and static.

5.5.1 External variables

Sometimes the source code for a C program is contained in more than one text file. If this
is the case, then it may be necessary to use variables that are defined in another file. You
can use a global variable in files other than the one in which it is defined by redeclaring it,
prefixed by the extern specifier, in the other files.
File main.c File secondary.c

#include <stdio.h>
int my_var;
int main()
{
extern int my_var; void print_value()
{
my_var = 500; printf("my_var = %d\n", my_var);
print_value(); }
exit (0);
}
In this example, the variable my_var is created in the file ‘secondary.c’, assigned a value
in the file ‘main.c’, and printed out in the function print_value, which is defined in the
file ‘secondary.c’, but called from the file ‘main.c’.
See Section 17.4 [Compiling multiple files], page 157, for information on how to compile
a program whose source code is split among multiple files. For this example, you can
simply type the command gcc -o testprog main.c secondary.c, and run the program
with ./testprog.

5.5.2 Static variables

A second important storage class specifier is static. Normally, when you call a function,
all its local variables are reinitialized each time the function is called. This means that
their values change between function calls. Static variables, however, maintain their value
between function calls.
Every global variable is defined as static automatically. (Roughly speaking, functions
anywhere in a program can refer to a global variable; in contrast, a function can only
refer to a local variable that is “nearby”, where “nearby” is defined in a specific manner.
See Chapter 6 [Scope], page 25, for more information on global variables. See Chapter 7
[Expressions and operators], page 29, for an example of a static local variable.)

5.5.3 Other storage classes

There are three more storage class identifiers in C: auto, register, and typedef.
• auto is the opposite of static. It is redundant, but is included in contemporary
versions of C for backwards compatibility. All local variables are auto by default.
• register is another outdated C storage class. Defining a variable as register used
to store it in one of the computer’s registers, a specific location on its processor chip,
Questions for Chapter 5 23

thereby making code using that variable run faster. These days, most C compilers
(including GCC) are smart enough to optimize the code (make it faster and more
compact) without the register keyword.
• typedef allows you to define your own variable types. See Chapter 19 [More data
types], page 177, for more information.

5.6 Questions for Chapter 5

1. What is an identifier?
2. Which of the following are valid C variable names?
1. Ralph23
2. 80shillings
3. mission_control
4. A%
5. A$
6. _off
3. Write a statement to declare two integers called start_temperature and
end_temperature.
4. What is the difference between the types float and double?
5. What is the difference between the types int and unsigned int?
6. Write a statement that assigns the value 1066 to the integer variable norman.
7. What data type do C functions return by default?
8. You must declare the data type a function returns at two places in a program. Where?
9. Write a statement, using the cast operator, to print out the integer part of the number
23.1256.
10. Is it possible to have an automatic global variable?
24 Chapter 5: Variables and declarations
Local Variables 25

6 Scope

Where a program’s fingers can and can’t reach.

Imagine that a function is a building with a person (Fred) standing in the doorway.
This person can see certain things: other people and other buildings, out in the open. But
Fred cannot see certain other things, such as the people inside the other buildings. Just so,
some variables in a C program, like the people standing outside, are visible to nearly every
other part of the program (these are called global variables), while other variables, like the
people indoors, are hidden behind the “brick walls” of curly brackets (these are called local
variables).
Where a variable is visible to other C code is called the scope of that variable. There
are two main kinds of scope, global and local, which stem from the two kinds of places in
which you can declare a variable:
1. Global scope is outside all of the functions, that is, in the space between function
definitions — after the #include lines, for example. Variables declared in global scope
are called global variables. Global variables can be used in any function, as well as in
any block within that function.
#include <stdio.h>

int global_integer;
float global_floating_point;

int main ()
{
exit (0);
}

2. You can also declare variables immediately following the opening bracket (‘{’) of any
block of code. This area is called local scope, and variables declared here are called
local variables. A local variable is visible within its own block and the ones that block
contains, but invisible outside its own block.
#include <stdio.h>

int main()
{
int foo;
float bar, bas, quux;

exit (0);
}

6.1 Global Variables

Global variables can be used in any function, as well as any block within that function.
(Technically, global variables can only be seen by functions that are defined after the dec-
laration of those global variables, but global variables are usually declared in a header file
that is included everywhere they are needed.) Global variables are created when a program
is started and are not destroyed until a program is stopped.
26 Chapter 6: Scope

6.2 Local Variables

Local variables, on the other hand, are only visible within local scope. They are “trapped”
inside their code blocks.
Just as global scope contains many functions, however, each function can contain many
code blocks (defined with curly brackets: ‘{...}’). C allows blocks within blocks, even
functions within functions, ad infinitum. A local variable is visible within its own block and
the ones that block contains, but invisible outside its own block.
int a;

/* Global scope. Global variable ’a’ is visible here,

but not local variables ’b’ or ’c’. */

int main()
{
int b;

/* Local scope of ’main’.

Variables ’a’ and ’b’ are visible here,
but not ’c’. */

{
int c;

/* Local scope of ‘{...}’ block within ’main’.

Variables ’a’, ’b’, and ’c’ are all visible here. */
}

exit (0);
}
Local variables are not visible outside their curly brackets. To use an “existence” rather
than a “visibility” metaphor, local variables are created when the opening brace is met, and
they are destroyed when the closing brace is met. (Do not take this too literally; they are
not created and destroyed in your C source code, but internally to the computer, when you
run the program.)

6.3 Communication via parameters

If no code inside a function could ever communicate with other parts of the program, then
functions would not be very useful. Functions would be isolated, comatose, unable to do
much of anything. Fortunately, although local variables are invisible outside their code
blocks, they can still communicate with other functions via parameters. See Chapter 7
[Expressions and operators], page 29, the next chapter, for information on parameters.

6.4 Scope example

Notice that there are two variables named my_var in the example below, both visible in the
same place. When two or more variables visible in one area of code have the same name,
the last variable to be defined takes priority. (Technically adept readers will realize that
this is because it was the last one onto the variable stack.)
/***************************************************************/
/* */
Questions for Chapter 6 27

/* SCOPE */
/* */
/***************************************************************/

#include <stdio.h>

int main ()
{
int my_var = 3;

{
int my_var = 5;
printf ("my_var=%d\n", my_var);
}

printf ("my_var=%d\n", my_var);

exit(0);
}
When you run this example, it will print out the following text:
my_var=5
my_var=3

6.5 Questions for Chapter 6

1. What is a global variable?
2. What is a local variable?
3. Do parameters spoil functions by leaking the variables into other functions?
4. Write a program gnoahs_park that declares 4 variables. Two global integer variables
called num_gnus and num_gnats, and two local floating point variables within the func-
tion main, called avg_gnu_mass, and avg_gnat_mass. Then add another function
called calculate_park_biomass, and pass avg_gnu_mass and avg_gnat_mass to it.
How many different storage spaces are used when this program runs? (Hint: are avg_
gnu_mass and avg_gnat_mass and their copies the same?)
28 Chapter 6: Scope
The assignment operator 29

7 Expressions and operators

Thinking in C. Short strings of code.
An operator is a character or string of characters used as a built-in function. We have
already experimented with one operator in C: the cast operator.
An operator is so called because it takes one or more values and operates on them to
produce a result. For example, the addition operator + can operate on the values 4 and 5
to produce the result 9. Such a procedure is called an operation, and any value operated
on (such as 4 and 5 in this example) is called an operand.
There are many operators in C. Some of them are familiar, such as the addition operator
+ and subtraction operator -. Most operators can be thought of as belonging to one of three
groups, according to what they do with their operands:
• Mathematical operators, such as the addition operator + in 100 + 500, or the multipli-
cation operator * in 12 * 2.
• Comparison operators (a subset of mathematical operators), such as the less-than op-
erator < and the greater-than operator >.
• Operators that produce new variable types, such as the cast operator.
The majority of operators fall into the first group. The second group is a subset of the
first set; in this second set, the result of an operation is a Boolean value (a value of either
true or false).
C has about forty different operators. The chief object of this chapter is to explain the
basic operators in C. We will examine more complex operators in another chapter. (See
Chapter 18 [Advanced operators], page 167.)

7.1 The assignment operator

No operator such as addition (+) or multiplication (*) would be useful without another
operator that attaches the values they produce to variables. Thus, the assignment operator
= is perhaps the most important mathematical operator.
We have seen the assignment operator already in our code examples. Here is an example
to refresh your memory:
int gnu_count, gnat_count, critter_count;

gnu_count = 45;
gnat_count = 5678;

critter_count = gnu_count + gnat_count;

The assignment operator takes the value of whatever is on the right-hand side of the =
symbol and puts it into the variable on the left-hand side. For example, the code sample
above assigns the value 45 to the variable gnu_count.
Something that can be assigned to is called an lvalue, (“l” for “left”, because it can
appear on the left side of an assignment). You will sometimes see the word ‘lvalue’ in
error messages from the compiler. For example, try to compile a program containing the
following code:
30 Chapter 7: Expressions and operators

5 = 2 + 3;
You will receive an error such as the following:
error
bad_example.c:3: invalid lvalue in assignment
You can’t assign a value to 5; it has its own value already! In other words, 5 is not an
lvalue.

7.1.1 Important note about assignment

Many people confuse the assignment operator (=) with the equality operator (==), and this
is a major source of bugs in C programs. Because of early arithmetic training, people tend
to think of = as indicating equality, but in C it means “takes on the value produced by”,
and it should always be read that way. By way of contrast, == is an equality test operator
and should always be read “is tested for equality with”. (See Section 7.8 [Comparisons and
logic], page 34, for more information on the == operator.)

7.2 Expressions and values

The most common operators in any language are basic arithmetic operators. In C, these
include the following:
+ unary plus, example: +5
- unary minus, example: -5
+ addition, example: 2 + 2
- subtraction, example: 14 - 7
* multiplication, example: 3 * 3
/ floating point division, example: 10.195 / 2.4
/ integer division div, example: 5 / 2
% integer remainder mod, example: 24 % 7

7.3 Expressions
An expression is simply a string of operators, variables, numbers, or some combination,
that can be parsed by the compiler. All of the following are expressions:
19

1 + 2 + 3

my_var

my_var + some_function()

(my_var + 4 * (some_function() + 2))

32 * circumference / 3.14

day_of_month % 7
Here is an example of some arithmetic expressions in C:
Parentheses and Priority 31

#include <stdio.h>

int main ()
{
int my_int;

printf ("Arithmetic Operators:\n\n");

my_int = 6;
printf ("my_int = %d, -my_int = %d\n", my_int, -my_int);

printf ("int 1 + 2 = %d\n", 1 + 2);

printf ("int 5 - 1 = %d\n", 5 - 1);
printf ("int 5 * 2 = %d\n", 5 * 2);

printf ("\n9 div 4 = 2 remainder 1:\n");

printf ("int 9 / 4 = %d\n", 9 / 4);
printf ("int 9 % 4 = %d\n", 9 % 4);

printf ("double 9 / 4 = %f\n", 9.0 / 4.0);

return 0;
}
The program above produces the output below:
Arithmetic Operators:

my_int = 6, -my_int = -6
int 1 + 2 = 3
int 5 - 1 = 4
int 5 * 2 = 10

9 div 4 = 2 remainder 1:
int 9 / 4 = 2
int 9 % 4 = 1
double 9 / 4 = 2.250000

7.4 Parentheses and Priority

Just as in algebra, the C compiler considers operators to have certain priorities, and eval-
uates, or parses, some operators before others. The order in which operators are evaluated
is called operator precedence or the order of operations. You can think of some operators
as “stronger” than others. The “stronger” ones will always be evaluated first; otherwise,
expressions are evaluated from left to right.
For example, since the multiplication operator * has a higher priority than the addition
operator + and is therefore evaluated first, the following expression will always evaluate to
10 rather than 18:
4 + 2 * 3
However, as in algebra, you can use parentheses to force the program to evaluate the
expression to 18:
(4 + 2) * 3
The parentheses force the expression (4 + 2) to be evaluated first. Placing parentheses
around 2 * 3, however, would have no effect.
32 Chapter 7: Expressions and operators

Parentheses are classed as operators by the compiler; they have a value, in the sense
that they assume the value of whatever is inside them. For example, the value of (5 + 5)
is 10.

7.5 Unary Operator Precedence

Unary operators are operators that have only a single operand — that is, they operate on
only one object. The following are (or can be) all unary operators:
++ -- + -
The order of evaluation of unary operators is from right to left, so an expression like:
*ptr++;
would perform the ++ before the *. (The ++ operator will be introduced in the next section,
and the * operator will be introduced in the next chapter. See Chapter 9 [Pointers], page 43.)

7.6 Special Assignment Operators ++ and --

C has some special operators that can simplify code. The simplest of these are the increment
and decrement operators:
++ increment: add one to
-- decrement: subtract one from
You can use these with any integer or floating point variable (or a character in some cases,
carefully). They simply add or subtract 1 from a variable. The following three statements
are equivalent:
variable = variable + 1;
variable++;
++variable;
So are these three:
variable = variable - 1;
variable--;
--variable;

Notice that the ++ and -- operators can be placed before or after the variable. In the
cases above, the two forms work identically, but there is actually a subtle difference. (See
Section 18.1.2 [Postfix and prefix ++ and –], page 168, for more information.)

7.7 More Special Assignments

Like ++ and --, the following operators are short ways of writing longer expressions. Con-
sider the following statement:
variable = variable + 23;
In C, this would be a long-winded way of adding 23 to variable. It could be done more
simply with the general increment operator +=, as in this example:
variable += 23;
This performs exactly the same operation. Similarly, the following two statements are
equivalent:
variable1 = variable1 + variable2;
variable1 += variable2;
There are a handful of these operators. For example, one for subtraction:
More Special Assignments 33

variable = variable - 42;

variable -= 42;
More surprisingly, perhaps, there is one for multiplication:
variable = variable * 2;
variable *= 2;

The main arithmetic operators all follow this pattern:

+= addition assignment operator
-= subtraction assignment operator
*= multiplication assignment operator
/= division assignment operator (floating point and integers)
%= remainder assignment operator (integers only)
There are more exotic kinds too, used for machine-level operations, which we will ignore
for the moment. (See Chapter 18 [Advanced operators], page 167, if you want to know
more.)
Here is a short program that demonstrates these special assignment operators:
#include <stdio.h>

int main()
{
int my_int;

printf ("Assignment Operators:\n\n");

my_int = 10; /* Assignment */

printf ("my_int = 10 : %d\n",my_int);

my_int++; /* my_int = my_int + 1 */

printf ("my_int++ : %d\n",my_int);

my_int += 5; /* my_int = my_int + 5 */

printf ("my_int += 5 : %d\n",my_int);

my_int--; /* my_int = my_int = 1 */

printf ("my_int-- : %d\n",my_int);

my_int -= 2; /* my_int = my_int - 2 */

printf ("my_int -= 2 : %d\n",my_int);

my_int *= 5; /* my_int = my_int * 5 */

printf ("my_int *= 5 : %d\n",my_int);

my_int /= 2; /* my_int = my_int / 2 */

printf ("my_int /= 2 : %d\n",my_int);

my_int %= 3; /* my_int = my_int % 3 */

printf ("my_int %%= 3 : %d\n",my_int);

return 0;
}
The program above produces the output below:
34 Chapter 7: Expressions and operators

Assignment Operators:

my_int = 10 : 10
my_int++ : 11
my_int += 5 : 16
my_int-- : 15
my_int -= 2 : 13
my_int *= 5 : 65
my_int /= 2 : 32
my_int %= 3 : 2
The second to last line of output is
my_int /= 2 : 32
In this example, 65 divided by 2 using the /= operator results in 32, not 32.5. This is
because both operands, 65 and 2, are integers, type int, and when /= operates on two
integers, it produces an integer result. This example only uses integer values, since that is
how the numbers are declared. To get the fractional answer, you would have had to declare
the three numbers involved as floats.
The last line of output is
my_int %= 3 : 2
This is because 32 divided by 3 is 10 with a remainder of 2.

7.8 Comparisons and logic

Comparison operators tell you how numerical values relate to one another, such as whether
they are equal to one another, or whether one is greater than the other. Comparison
operators are used in logical tests, such as if statements. (See Chapter 10 [Decisions],
page 51.)
The results of a logical comparison are always either true (1) or false (0). In computer
programming jargon, true and false are the two Boolean values. Note that, unlike real life,
there are no “gray areas” in C; just as in Aristotelian logic, a comparison operator will
never produce a value other than true or false.
Six operators in C are used to make logical comparisons:
== is equal to
!= is not equal to
> is greater than
< is less than
>= is greater than or equal to
<= is less than or equal to
Important: Remember that many people confuse the equality operator (==) with the as-
signment operator (=), and this is a major source of bugs in C programs. (See Section 7.2
[Expressions and values], page 30, for more information on the distinction between the ==
and = operators.)
The operators above result in values, much as the addition operator + does. They
produce Boolean values: true and false only. Actually, C uses 1 for “true” and 0 for “false”
Inclusive OR 35

when evaluating expressions containing comparison operators, but it is easy to define the
strings ‘TRUE’ and ‘FALSE’ as macros, and they may well already be defined in a library
file you are using. (See Chapter 12 [Preprocessor directives], page 67, for information on
defining macros.)
#define TRUE 1
#define FALSE 0
Note that although any non-zero value in C is treated as true, you do not need to worry
about a comparison evaluating to anything other than 1 or 0. Try the following short
program:
#include <stdio.h>

int main ()
{
int truth, falsehood;

truth = (2 + 2 == 4);
falsehood = (2 + 2 == 5);

printf("truth is %d\n", truth);

printf("falsehood is %d\n", falsehood);

exit (0);
}
You should receive the following result:
truth is 1
falsehood is 0

7.9 Logical operators

Comparisons are often made in pairs or groups. For example, you might want to ask a
question such as, “Is variable a greater than variable b and is variable b greater than
variable c?” The word “and” in the question above is represented in C by the logical
operator (an “operator on Boolean values”) &&, and the whole comparison above might be
represented by the following expression:
(a > b) && (b > c)
The main logical operators in C are as follows:
&& logical AND
|| logical Inclusive OR (See Section 7.9.1 [Inclusive OR], page 35.)
! logical NOT
Here is another example. The question, “Is the variable a greater than the variable b,
or is the variable a not greater than the variable c?” might be written:
(a > b) || !(a > c)

7.9.1 Inclusive OR
Note well! Shakespeare might have been disappointed that, whatever the value of a variable
to_be, the result of
36 Chapter 7: Expressions and operators

to_be || !to_be
(i.e. “To be, or not to be?”) is always 1, or true. This is because one or the other of to_be
or !to_be must always be true, and as long as one side of an OR || expression is true, the
whole expression is true.

7.10 Questions for Chapter 7

1. What is an operand?
2. Write a short statement that assigns the remainder of 5 divided by 2 to a variable
called remainder and prints it out.
3. Write a statement that subtracts -5 from 10.
Parameters 37

8 Parameters
Ways in and out of functions.
Parameters are the main way in C to transfer, or pass, information from function to
function. Consider a call to our old friend calculate_bill:
total = calculate_bill (20, 35, 28);
We are passing 20, 35, and 28 as parameters to calculate_bill so that it can add them
together and return the sum.
When you pass information to a function with parameters, in some cases the information
can go only one way, and the function returns only a single value (such as total in the
above snippet of code). In other cases, the information in the parameters can go both ways;
that is, the function called can alter the information in the parameters it is passed.
The former technique (passing information only one way) is called passing parameters
by value in computer programming jargon, and the latter technique (passing information
both ways) is referred to as passing parameters by reference.
For our purposes, at the moment, there are two (mutually exclusive) kinds of parameters:
• Value parameters are the kind that pass information one-way. They are so-called
because the function to which they are passed receives only a copy of their values,
and they cannot be altered as variable parameters can. The phrase “passing by value”
mentioned above is another way to talk about passing “value parameters”.
• Variable parameters are the kind that pass information back to the calling function.
They are so called because the function to which they are passed can alter them, just as
it can alter an ordinary variable. The phrase “passing by reference” mentioned above
is another way to talk about passing “variable parameters”.
Consider a slightly-expanded version of calculate_bill:
#include <stdio.h>

int main (void);

int calculate_bill (int, int, int);

int main()
{
int bill;
int fred = 25;
int frank = 32;
int franny = 27;

bill = calculate_bill (fred, frank, franny);

printf("The total bill comes to $%d.00.\n", bill);

exit (0);
}

int calculate_bill (int diner1, int diner2, int diner3)

{
int total;

total = diner1 + diner2 + diner3;

return total;
38 Chapter 8: Parameters

}
Note that all of the parameters in this example are value parameters: the information
flows only one way. The values are passed to the function calculate_bill. The original
values are not changed. In slightly different jargon, we are “passing the parameters by value
only”. We are not passing them “by reference”; they are not “variable parameters”.
All parameters must have their types declared. This is true whether they are value
parameters or variable parameters. In the function calculate_bill above, the value pa-
rameters diner1, diner2, and diner3 are all declared to be of type int.

8.1 Parameters in function prototypes

Note that in the function prototype for calculate_bill, the parameter names were com-
pletely omitted. This is perfectly acceptable in ANSI C, although it might be confusing to
someone trying to understand your code by reading the function prototypes, which can be
in a separate file from the functions themselves. For instance, in the code example above,
the function prototype for calculate_bill looks like this:
int calculate_bill (int, int, int);
You may include parameter names in function prototypes if you wish; this is usually a
good idea when the function prototype is significantly separated from the function definition,
such as when the prototype is in a header file or at the top of a long file of function definitions.
For example, we could have written the prototype for calculate_bill thus:
int calculate_bill (int diner1, int diner2, int diner3);
Parameter names in a function prototype do not need to match the names in the func-
tion’s definition; only their types need to match. For example, we can also write the function
prototype above in this way:
int calculate_bill (int guest1, int guest2, int guest3);
As usual, it is a good idea to use mnemonic names for the parameters in a function
prototype, as in the last two examples.1 Thus, the function prototype below is not as
helpful to the person reading your code as the last two examples are; it might just as well
have been written without variable names at all:
int calculate_bill (int variable1, int variable2, int variable3);

8.2 Value Parameters

When you are passing data to a function by value, the parameters in the function you are
passing the data to contain copies of the data in the parameters you are passing the data
with. Let us modify the function main from the last example slightly:
int main()
{
int bill;
int fred = 25;
int frank = 32;
int franny = 27;

bill = calculate_bill (fred, frank, franny);

1
That is, unless you are competing in The International Obfuscated C Code Contest.
Actual parameters and formal parameters 39

fred = 20000;
frank = 50000;
franny = 20000;

printf("The total bill comes to $%d.00.\n", bill);

exit (0);
}
As far as the function calculate_bill is concerned, fred, frank, and franny are still
25, 32, and 27 respectively. Changing their values to extortionate sums after passing them
to calculate_bill does nothing; calculate_bill has already created local copies of the
parameters, called diner1, diner2, and diner3 containing the earlier values.
Important: Even if we named the parameters in the definition of calculate_bill to
match the parameters of the function call in main (see example below), the result would be
the same: main would print out ‘$84.00’, not ‘$90000.00’. When passing data by value,
the parameters in the function call and the parameters in the function definition (which are
only copies of the parameters in the function call) are completely separate.
Just to remind you, this is the calculate_bill function:
int calculate_bill (int fred, int frank, int franny)
{
int total;

total = fred + frank + franny;

return total;
}

8.3 Actual parameters and formal parameters

There are two other categories that you should know about that are also referred to as
“parameters”. They are called “parameters” because they define information that is passed
to a function.
• Actual parameters are parameters as they appear in function calls.
• Formal parameters are parameters as they appear in function declarations.
A parameter cannot be both a formal and an actual parameter, but both formal param-
eters and actual parameters can be either value parameters or variable parameters.
Let’s look at calculate_bill again:
#include <stdio.h>

int main (void);

int calculate_bill (int, int, int);

int main()
{
int bill;
int fred = 25;
int frank = 32;
int franny = 27;

bill = calculate_bill (fred, frank, franny);

printf("The total bill comes to $%d.00.\n", bill);
40 Chapter 8: Parameters

exit (0);
}

int calculate_bill (int diner1, int diner2, int diner3)

{
int total;

total = diner1 + diner2 + diner3;

return total;
}

In the function main in the example above, fred, frank, and franny are all actual
parameters when used to call calculate_bill. On the other hand, the corresponding
variables in calculate_bill (namely diner1, diner2 and diner3, respectively) are all
formal parameters because they appear in a function definition.
Although formal parameters are always variables (which does not mean that they are
always variable parameters), actual parameters do not have to be variables. You can use
numbers, expressions, or even function calls as actual parameters. Here are some examples
of valid actual parameters in the function call to calculate_bill:
bill = calculate_bill (25, 32, 27);

bill = calculate_bill (50+60, 25*2, 100-75);

bill = calculate_bill (fred, franny, (int) sqrt(25));

(The last example requires the inclusion of the math routines in ‘math.h’, and compilation
with the ‘-lm’ option. sqrt is the square-root function and returns a double, so it must be
cast into an int to be passed to calculate_bill.)

8.4 Variadic functions

Suppose you are writing a program that repeatedly generates lists of numbers that can
run anywhere from one to fifty items. You never know how many numbers a particular
list will contain, but you always want to add all the numbers together. Passing them to
an ordinary C function will not work, because an ordinary function has a fixed number of
formal parameters, and cannot accept an arbitrarily long list of actual parameters. What
should you do?
One way of solving this problem is to use a variadic function, or function that can accept
arbitrarily long lists of actual parameters. You can do this by including the ‘stdarg.h’
header in your program. For example, with ‘stdarg.h’, you can write a function called
add_all that will add all integers passed to it, returning correct results for all of the
following calls:
sum = add_all (2, 3, 4);

sum = add_all (10, 150, 9, 81, 14, 2, 2, 31);

sum = add_all (4);

Unfortunately, the use of ‘stdarg.h’ is beyond the scope of this tutorial. For more
information on variadic functions, see the GNU C Library manual.
Questions for Chapter 8 41

8.5 Questions for Chapter 8

1. What is the difference between a value parameter and a variable parameter?
2. What is the difference between a formal parameter and an actual parameter?
3. What does passing by reference let you do that passing by value doesn’t?
4. Can a function call be used as an actual parameter?
5. Do actual and formal parameters need to have the same names?
42 Chapter 8: Parameters
Pointer operators 43

9 Pointers
Making maps of data.
In one sense, any variable in C is just a convenient label for a chunk of the computer’s
memory that contains the variable’s data. A pointer, then, is a special kind of variable that
contains the location or address of that chunk of memory. (Pointers are so called because
they point to a chunk of memory.) The address contained by a pointer is a lengthy number
that enables you to pinpoint exactly where in the computer’s memory the variable resides.
Pointers are one of the more versatile features of C. There are many good reasons to use
them. Knowing a variable’s address in memory enables you to pass the variable to a function
by reference (See Section 9.4 [Variable parameters], page 47.)1 Also, since functions are just
chunks of code in the computer’s memory, and each of them has its own address, you can
create pointers to functions too, and knowing a function’s address in memory enables you to
pass functions as parameters too, giving your functions the ability to switch among calling
numerous functions. (See [Function pointers], page 248.)
Pointers are important when using text strings. In C, a text string is always accessed
with a pointer to a character — the first character of the text string. For example, the
following code will print the text string ‘Boy howdy!’:
char *greeting = "Boy howdy!";
printf ("%s\n\n", greeting);
See Chapter 15 [Strings], page 97.
Pointers are important for more advanced types of data as well. For example, there
is a data structure called a “linked list” that uses pointers to “glue” the items in the list
together. (See Chapter 20 [Data structures], page 183, for information on linked lists.)
Another use for pointers stems from functions like the C input routine scanf. This
function accepts information from the keyboard, just as printf sends output to the console.
However, scanf uses pointers to variables, not variables themselves. For example, the
following code reads an integer from the keyboard:
int my_integer;
scanf ("%d", &my_integer);
(See Section 16.2.9.1 [scanf], page 128, for more information.)

9.1 Pointer operators

To create a pointer to a variable, we use the * and & operators. (In context, these have
nothing to do with multiplication or logical AND. For example, the following code declares
a variable called total_cost and a pointer to it called total_cost_ptr.
float total_cost;
float *total_cost_ptr;

total_cost_ptr = &total_cost;
The ‘*’ symbol in the declaration of total_cost_ptr is the way to declare that variable
to be a pointer in C. (The ‘_ptr’ at the end of the variable name, on the other hand, is just
a way of reminding humans that the variable is a pointer.)
1
This, by the way, is how the phrase “pass by reference” entered the jargon. Like other pointers, a variable
parameter “makes a reference” to the address of a variable.
44 Chapter 9: Pointers

When you read C code to yourself, it is often useful to be able to pronounce C’s operators
aloud; you will find it can help you make sense of a difficult piece of code. For example,
you can pronounce the above statement float *total_cost_ptr as “Declare a float pointer
called total_cost_ptr”, and you can pronounce the statement total_cost_ptr = &total_
cost; as “Let total_cost_ptr take as its value the address of the variable total_cost”.
Here are some suggestions for pronouncing the * and & operators, which are always
written in front of a variable:
* “The contents of the address held in variable” or “the contents of the location
pointed to by variable”.
& “The address of variable” or “the address at which the variable variable is
stored”.
For instance:
&fred “The address of fred” or “the address at which the variable fred is stored”.
*fred_ptr
“The contents of the address held in fred_ptr” or “the contents of the location
pointed to by fred_ptr”.
The following examples show some common ways in which you might use the * and &
operators:
int some_var; /* 1 */
“Declare an integer variable called some_var.”

int *ptr_to_some_var; /* 2 */
“Declare an integer pointer called ptr_to_some_var.” (The
* in front of ptr_to_some_var is the way C declares
ptr_to_some_var as a pointer to an integer, rather than just an
integer.)

some_var = 42; /* 3 */
“Let some_var take the value 42.”

ptr_to_some_var = &some_var; /* 4 */
“Let ptr_to_some_var take the address of the variable
some_var as its value.” (Notice that only now does
ptr_to_some_var become a pointer to the particular variable
some_var — before this, it was merely a pointer that could
point to any integer variable.)

printf ("%d\n\n", *ptr_to_some_var); /* 5 */

“Print out the contents of the location pointed to by
ptr_to_some_var.” (In other words, print out some_var
itself. This will print just 42. Accessing what a pointer points to in
this way is called dereferencing the pointer, because the pointer
is considered to be referencing the variable.)

*ptr_to_some_var = 56; /* 6 */ “Let the contents of the location

pointed to by ptr_to_some_var equal 56.” (In the context of the
other statements, this is the same as the more direct statement
some_var = 56;.)
A subtle point: don’t confuse the usage of asterisks in code like examples 2 and 6 above.
Using an asterisk in a declaration, as in example 2, declares the variable to be a pointer,
Pointer types 45

while using it on the left-hand side of an assignment, as in example 6, dereferences a variable

that is already a pointer, enabling you to access the variable to which the pointer is pointing.

9.2 Pointer types

Pointers can point to any type of variable, but they must be declared to do so. A pointer
to an integer is not the same type of variable as a pointer to a float or other variable type.
At the “business end” of a pointer is usually a variable, and all variables have a type.
Here are some examples of different types of pointer:
int *my_integer_ptr;

char *my_character_ptr;

float *my_float_ptr;

double *my_double_ptr;
However, GCC is fairly lenient about casting different types of pointer to one another
implicitly, or automatically, without your intervention. For example, the following code will
simply truncate the value of *float_ptr and print out 23. (As a bonus, pronunciation is
given for every significant line of the code in this example.)
#include <stdio.h>
/* Include the standard input/output header in this program */

int main()
/* Declare a function called main that returns an integer
and takes no parameters */
{
int *integer_ptr;
/* Declare an integer pointer called integer_ptr */

float *float_ptr;
/* Declare a floating-point pointer called float_ptr */

int my_int = 17;

/* Declare an integer variable called my_int
and assign it the value 17 */

float my_float = 23.5;

/* Declare a floating-point variable called my_float
and assign it the value 23.5 */

integer_ptr = &my_int;
/* Assign the address of the integer variable my_int
to the integer pointer variable integer_ptr */

float_ptr = &my_float;
/* Assign the address of the floating-point variable my_float
to the floating-point pointer variable float_ptr */

*integer_ptr = *float_ptr;
/* Assign the contents of the location pointed to by
the floating-point pointer variable float_ptr
to the location pointed to by the integer pointer variable
integer_ptr (the value assigned will be truncated) */
46 Chapter 9: Pointers

printf ("%d\n\n", *integer_ptr);

/* Print the contents of the location pointed to by the
integer pointer variable integer_ptr */

return 0;
/* Return a value of 0, indicating successful execution,
to the operating system */

There will still be times when you will want to convert one type of pointer into another.
For example, GCC will give a warning if you try to pass float pointers to a function that
accepts integer pointers. Not treating pointer types interchangeably will also help you
understand your own code better.
To convert pointer types, use the cast operator. (See Section 5.4 [The cast operator],
page 20.) As you know, the general form of the cast operator is as follows:
(type ) variable

Here is the general form of the cast operator for pointers:

(type *) pointer_variable

Here is an actual example:

int *my_integer_ptr;
long *my_long_ptr;

my_long_ptr = (long *) my_integer_ptr;

This copies the value of the pointer my_integer to the pointer my_long_ptr. The cast
operator ensures that the data types match. (See Chapter 20 [Data structures], page 183,
for more details on pointer casting.)

9.3 Pointers and initialization

You should not initialize pointers with a value when you declare them, although the compiler
will not prevent this. Doing so simply makes no sense. For example, think about what
happens in the following statement:
int *my_int_ptr = 2;

First, the program allocates space for a pointer to an integer. Initially, the space will contain
garbage (random data). It will not contain actual data until the pointer is “pointed at”
such data. To cause the pointer to refer to a real variable, you need another statement,
such as the following:
my_int_ptr = &my_int;

On the other hand, if you use just the single initial assignment, int *my_int_ptr = 2;,
the program will try to fill the contents of the memory location pointed to by my_int_ptr
with the value 2. Since my_int_ptr is filled with garbage, it can be any address. This
means that the value 2 might be stored anywhere. anywhere, and if it overwrites something
important, it may cause the program to crash.
The compiler will warn you against this. Heed the warning!
Passing pointers correctly 47

9.4 Variable parameters

Now that you know something about pointers, we can discuss variable parameters and
passing by reference in more detail. (See Chapter 8 [Parameters], page 37, to refresh your
memory on this topic.)
There are two main ways to return information from a function. The most common way
uses the return command. However, return can only pass one value at a time back to the
calling function. The second way to return information to a function uses variable param-
eters. Variable parameters (“passing by reference”) enable you to pass back an arbitrary
number of values, as in the following example:
#include <stdio.h>

int main();
void get_values (int *, int *);

int main()
{
int num1, num2;
get_values (&num1, &num2);

printf ("num1 = %d and num2 = %d\n\n", num1, num2);

return 0;
}

void get_values (int *num_ptr1, int *num_ptr2)

{
*num_ptr1 = 10;
*num_ptr2 = 20;
}
The output from this program reads:
num1 = 10 and num2 = 20
Note that we do use a return command in this example — in the main function. Remember,
main must always be declared of type int and should always return an integer value. (See
Chapter 22 [Style], page 203.)
When you use value parameters, the formal parameters (the parameters in the function
being called) are mere copies of the actual parameters (the parameters in the function call).
When you use variable parameters, on the other hand, you are passing the addresses of
the variables themselves. Therefore, in the program above, it is not copies of the variables
num1 and num2 that are passed to get_values, but the addresses of their actual memory
locations. This information can be used to alter the variables directly, and to return the
new values.

9.4.1 Passing pointers correctly

You might be wondering why main calls the function get_values above with ampersands
before the parameters —
get_values (&num1, &num2);
— while the function itself is defined with asterisks before its parameters:
void get_values (int *num_ptr1, int *num_ptr2)
48 Chapter 9: Pointers

{
*num_ptr1 = 10;
*num_ptr2 = 20;
}
Think carefully for a moment about what is happening in these fragments of code. The
variables num1 and num2 in main are ordinary integers, so when main prefixes them with
ampersands (&) while passing them to get_values, it is really passing integer pointers.
Remember, &num1 should be read as “the address of the variable num1”.
The code reads like this:
get_values (&num1, &num2);

“Evaluate the function get_values, passing to it the

addresses at which the variables num1 and num2 are stored.”.
The function get_values is defined like this:
void get_values (int *num_ptr1, int *num_ptr2)

“Define the function get_values. It returns a void

value (so it operates only via “side effects” on the variable
parameters it is passed). It takes two parameters, both of type
int *. The first parameter is called num_ptr1 and is a
pointer to an integer value, and the second parameter is called
num_ptr2 and is also a pointer to an integer value. When this
function is called, it must be passed the addresses of variables, not the
variables themselves.”
Remember that declaring a variable with an asterisk (*) before it means “declare this
variable to be a pointer”, so the formal parameters of get_values are integer pointers. The
parameters must be declared this way, because the main function sends the addresses of
num1 and num2 — that is, by the time the get_values function receives the parameters,
they are already pointers — hence their names in get_values: num_ptr1 and num_ptr2,
rather than num1 and num2.
In effect, we are “matching up” the data types of num1 and num2 with those of num_ptr1
and num_ptr2, respectively, when we prefix num1 and num2 with ampersands while passing
them, and prefix num_ptr1 and num_ptr2 with asterisks in the parameter list of the function
get_values. We do not have to write num_ptr1 = &num1; and num_ptr2 = &num2; — the
calling convention does that for us.
Important! This is a general rule in C: when you pass actual parameters as pointers
using ampersands (e.g. &num1, “the address of the variable num1”), you must use asterisks
to declare as pointers the corresponding formal parameters in the function to which you
pass them, (e.g. int *num_ptr1, “the contents of the location pointed to by num_ptr1”).

9.4.2 Another variable parameter example

There is nothing mysterious about pointers, but they can be tricky. Here is another example.
Notice that the pointers in both this example and the example above are dereferenced
with asterisks before they are used (for instance, when the contents of the location pointed to
by height_ptr are multiplied by the integer hscale with the line *height_ptr = *height_
ptr * hscale; in the function scale_dimensions below).
#include <stdio.h>
Questions for Chapter 9 49

int main();
void scale_dimensions (int *, int *);

/* Scale some measurements */

int main()
{
int height,width;

height = 4;
width = 5;

scale_dimensions (&height, &width);

printf ("Scaled height = %d\n", height);

printf ("Scaled width = %d\n", width);

return 0;
}

void scale_dimensions (int *height_ptr, int *width_ptr)

{
int hscale = 3; /* scale factors */
int wscale = 5;

*height_ptr = *height_ptr * hscale;

*width_ptr = *width_ptr * wscale;
}

9.5 Questions for Chapter 9

1. What is a pointer?
2. How is a variable declared to be a pointer?
3. What data types can pointers point to?
4. Write a statement which converts a pointer to an integer into a pointer to a double
type.
5. Why is it incorrect to write float *number = 2.65; ?
50 Chapter 9: Pointers
if 51

10 Decisions
Testing and Branching. Making conditions.
Until now, our code examples have been linear: control has flowed in one direction from
start to finish. In this chapter, we will examine ways to enable code to make decisions
and to choose among options. You will learn how to program code that will function in
situations similar to the following:
• If the user hits the jackpot, print a message to say so: ‘You’ve won!’
• If a bank balance is positive, then print ‘C’ for “credit”; otherwise, print ‘D’ for “debit”.
• If the user has typed in one of five choices, then do something that corresponds to the
choice, otherwise display an error message.
In the first case there is a simple “do or don’t” choice. In the second case, there are two
choices. The final case contains several possibilities.
C offers four main ways of coding decisions like the ones above. They are listed below.
if...
if (condition )
{
do something
}

if...else...
if (condition )
{
do something
}
else
{
do something else
}

...?...:...
(condition ) ? do something : do something else ;
switch
switch (condition )
{
case first case : do first thing
case second case : do second thing
case third case : do third thing
}

10.1 if
The first form of the if statement is an all-or-nothing choice: if some condition is satisfied,
do something; otherwise, do nothing. For example:
if (condition ) statement ;
or
if (condition )
{
compound statement
}
52 Chapter 10: Decisions

In the second example, instead of a single statement, a whole block of statements is

executed. In fact, wherever you can place a single statement in C, you can place a compound
statement instead: a block of statements enclosed by curly brackets.
A condition is usually an expression that makes some sort of comparison. It must be
either true or false, and it must be enclosed in parentheses: ‘(...)’. If the condition is
true, then the statement or compound statement following the condition will be executed;
otherwise, it will be ignored. For example:
if (my_num == 0)
{
printf ("The number is zero.\n");
}

if (my_num > 0)
{
printf ("The number is positive.\n");
}

if (my_num < 0)
{
printf ("The number is negative.\n");
}
The same code could be written more compactly in the following way:
if (my_num == 0) printf ("The number is zero.\n");
if (my_num > 0) printf ("The number is positive.\n");
if (my_num < 0) printf ("The number is negative.\n");
It is often a good idea stylistically to use curly brackets in an if statement. It is no
less efficient from the compiler’s viewpoint, and sometimes you will want to include more
statements later. It also makes if statements stand out clearly in the code. However, curly
brackets make no sense for short statements such as the following:
if (my_num == 0) my_num++;
The if command by itself permits only limited decisions. With the addition of else in
the next section, however, if becomes much more flexible.

10.2 if... else...

Let’s review the basic form of the if... else... statement:
if (condition )
{
compound statement
}
else
{
compound statement
}
As with the bare if statement, there is a simplified version of the if... else... state-
ment without code blocks:
if (condition ) statement else statement ;
When the if... else... is executed, the condition in parentheses is evaluated. If it is
true, then the first statement or code block is executed; otherwise, the second statement or
code block is executed. This can save unnecessary tests and make a program more efficient:
Nested if statements 53

if (my_num > 0)
{
printf ("The number is positive.");
}
else
{
printf ("The number is zero or negative.");
}
It is not necessary to test my_num in the second block because that block is not executed
unless my_num is not greater than zero.

10.3 Nested if statements

Consider the following two code examples. Their purposes are exactly the same.
int my_num = 3;

if ((my_num > 2) && (my_num < 4))

{
printf ("my_num is three");
}
or:
int my_num =3;

if (my_num > 2)
{
if (my_num < 4)
{
printf ("my_num is three");
}
}
Both of these code examples have the same result, but they arrive at it in different ways.
The first example, when translated into English, might read, “If my_num is greater than
two and my_num is less than four (and my_num is an integer), then my_num has to be three.”
The second method is more complicated. In English, it can be read, “If my_num is greater
than two, do what is in the first code block. Inside it, my_num is always greater than two;
otherwise the program would never have arrived there. Now, if my_num is also less than
four, then do what is inside the second code block. Inside that block, my_num is always less
than four. We also know it is more than two, since the whole of the second test happens
inside the block where that’s true. So, assuming my_num is an integer, it must be three.”
In short, there are two ways of making compound decisions in C. You make nested tests,
or you can use the comparison operators &&, ||, and so on. In situations where sequences of
comparison operators become too complex, nested tests are often a more attractive option.
Consider the following example:
if (i > 2)
{
/* i is greater than 2 here! */
}
else
{
/* i is less than or equal to 2 here! */
}
54 Chapter 10: Decisions

The code blocks in this example provide “safe zones” wherein you can rest assured that
certain conditions hold. This enables you to think and code in a structured way.
You can nest if statements in multiple levels, as in the following example:
#include <stdio.h>

int main ()
{
int grade;

printf("Type in your grade: ");

scanf ("%d", &grade);

if (grade < 10)

{
printf ("Man, you’re lame! Just go away.\n");
}
else
{
if (grade < 65)
{
printf ("You failed.\n");
}
else
{
printf ("You passed!\n");
if (grade >= 90)
{
printf ("And you got an A!\n");
}
else
{
printf ("But you didn’t get an A. Sorry.\n");
}
}
}
return 0;
}

10.4 The ?. . . :. . . operator

The ?. . . :. . . operator is a sort of shorthand if. . . else. . . statement. Because it is a little
cryptic, it is not often used, but the basic form is as follows:
(condition ) ? expression1 : expression2 ;
The program evaluates condition. If it is true (not zero), then expression1 is returned;
otherwise, expression2 is returned.
For example, in the short program below, the line bas = (foo > bar) ? foo : bar; as-
signs foo to bas if foo is greater than bar; otherwise, it assigns bar to bas.
#include <stdio.h>

int main()
{
int foo = 10;
int bar = 50;
int bas;
Example Listing 55

bas = (foo > bar) ? foo : bar;

printf("bas = %d\n\n", bas);

return 0;
}
The program will print ‘bas = 50’ as a result.

10.5 The switch statement

The switch construction is another way of making decisions in C code. It is very flexible,
but only tests for integer and character values. It has the following general form:
switch (integer or character expression )
{
case constant1 : statement1 ;
break; /* optional */

case constant2 : statement2 ;

break; /* optional */

case constant3 : statement3 ;

break; /* optional */
...
}
The integer or character expression in the parentheses is evaluated, and the program checks
whether it matches one of the constants in the various cases listed. If there is a match, the
statement following that case will be executed, and execution will continue until either a
break statement or the closing curly bracket of the entire switch statement is encountered.
One of the cases is called default. Statements after the default case are executed
when none of the other cases are satisfied. You only need a default case if you are not sure
you are covering every case with the ones you list.
Here is an example program that uses the switch statement to translate decimal digits
into Morse code:

10.6 Example Listing

#include <stdio.h>

int main ();

void morse (int);

int main ()
{
int digit;

printf ("Enter any digit in the range 0 to 9: ");

scanf ("%d", &digit);

if ((digit < 0) || (digit > 9))

{
printf ("Your number was not in the range 0 to 9.\n");
}
else
56 Chapter 10: Decisions

{
printf ("The Morse code of that digit is ");
morse (digit);
}
return 0;
}

void morse (int digit) /* print out Morse code */

{
switch (digit)
{
case 0 : printf ("-----");
break;
case 1 : printf (".----");
break;
case 2 : printf ("..---");
break;
case 3 : printf ("...--");
break;
case 4 : printf ("....-");
break;
case 5 : printf (".....");
break;
case 6 : printf ("-....");
break;
case 7 : printf ("--...");
break;
case 8 : printf ("---..");
break;
case 9 : printf ("----.");
}
printf ("\n\n");
}
The morse function selects one of the printf statements with switch, based on the integer
expression digit. After every case in the switch, a break statement is used to jump switch
statement’s closing bracket ‘}’. Without break, execution would fall through to the next
case and execute its printf statement.
Here is an example of using fallthrough in a constructive way. The function yes accepts
input from the user and tests whether it was ’y’ or ’Y’. (The getchar function is from
the standard library and reads a character of input from the terminal. See Section 16.3.1
[getchar], page 130.)
#include <stdio.h>

int main ()
{
printf ("Will you join the Free Software movement? ");
if (yes())
{
printf("Great! The price of freedom is eternal vigilance!\n\n");
}
else
{
printf("Too bad. Maybe next life...\n\n");
}
Questions for Chapter 10 57

return 0;
}

int yes()
{
switch (getchar())
{
case ’y’ :
case ’Y’ : return 1;
default : return 0;
}
}
If the character is ‘y’, then the program falls through and meets the statement return 1. If
there were a break statement after case ’y’, then the program would not be able to reach
case ’Y’ unless an actual ‘Y’ were typed.
Note: The return statements substitute for break in the above code, but they do more
than break out of switch — they break out of the whole function. This can be a useful
trick.

10.7 Questions for Chapter 10

1. Translate the following into good C: “If 1 does not equal 42, print out ‘Thank heavens
for mathematics!’ ”
2. Write a program to get a lot of numbers from the user and print out the maximum and
minimum of those.
3. Write an automatic teller machine program that simulates telling you your bank balance
when you enter your account number and PIN number, but otherwise displays an error.
4. Write a mock program for a car computer that tells you how many kilometers to the
liter you’re getting when you enter how many liters of gas you’ve used and how far you
travelled.
58 Chapter 10: Decisions
do. . . while 59

11 Loops
Controlling repetitive processes. Nesting loops
Loops are a kind of C construct that enable the programmer to execute a sequence of
instructions over and over, with some condition specifying when they will stop. There are
three kinds of loop in C:
• while
• do . . . while
• for

11.1 while
The simplest of the three is the while loop. It looks like this:
while (condition )
{
do something
}
The condition (for example, (a > b)) is evaluated every time the loop is executed. If
the condition is true, then statements in the curly brackets are executed. If the condition
is false, then those statements are ignored, and the while loop ends. The program then
executes the next statement in the program.
The condition comes at the start of the loop, so it is tested at the start of every pass,
or time through the loop. If the condition is false before the loop has been executed even
once, then the statements inside the curly brackets will never be executed. (See Section 11.2
[do...while], page 60, for an example of a loop construction where this is not true.)
The following example prompts the user to type in a line of text, and then counts all the
spaces in the line. The loop terminates when the user hits the hRETi key and then prints
out the number of spaces. (See Section 16.3.1 [getchar], page 130, for more information on
the standard library getchar function.)
#include <stdio.h>

int main()
{
char ch;
int count = 0;

printf ("Type in a line of text.\n");

while ((ch = getchar()) != ’\n’)

{
if (ch == ’ ’)
{
count++;
}
}

printf ("Number of spaces = %d.\n\n", count);

return 0;
}
60 Chapter 11: Loops

11.2 do. . . while

The do..while loop has the form:
do
{
do something
}
while (condition );
Notice that the condition is at the end of this loop. This means that a do..while loop
will always be executed at least once, before the test is made to determine whether it should
continue. This is the chief difference between while and do. . . while.
The following program accepts a line of input from the user. If the line contains a string
of characters delimited with double quotation marks, such as ‘"Hello!"’, the program prints
the string, with quotation marks. For example, if the user types in the following string:
I walked into a red sandstone building. "Oof!" [Careful, Nick!]
. . . then the program will print the following string:
"Oof!"
If the line contains only one double quotation mark, then the program will display an error,
and if it contains no double quotation marks, the program will print nothing.
Notice that the do. . . while loop in main waits to detect a linefeed character (\n), while
the one in get_substring looks for a double quotation mark (‘"’), but checks for a linefeed
in the loop body, or main code block of the loop, so that it can exit the loop if the user
entered a linefeed prematurely (before the second ‘"’).
This is one of the more complex examples we have examined so far, and you might find
it useful to trace the code, or follow through it step by step.
#include <stdio.h>

int main();
void get_substring();

int main()
{
char ch;

printf ("Enter a string with a quoted substring:\n\n");

do
{
ch = getchar();
if (ch == ’"’)
{
putchar(ch);
get_substring();
}
}
while (ch != ’\n’);

return 0;
}
for 61

void get_substring()
{
char ch;

do
{
ch = getchar();
putchar(ch);

if (ch == ’\n’)
{
printf ("\nString was not closed ");
printf ("before end of line.\n");
break;
}
}
while (ch != ’"’);

printf ("\n\n");
}

11.3 for
The most complex loop in C is the for loop. The for construct, as it was developed
in earlier computer languages such as BASIC and Pascal, was intended to behave in the
following way:
For all values of variable from value1 to value2, in steps of value3, repeat the
following sequence of commands. . .
The for loop in C is much more versatile than its counterpart in those earlier languages.
The for loop looks like this in C:
for (initialization ; condition ; increment )
{
do something ;
}
In normal usage, these expressions have the following significance.
• initialization
This is an expression that initializes the control variable, or the variable tested in
the condition part of the for statement. (Sometimes this variable is called the loop’s
index.) The initialization part is only carried out once before the start of the loop.
Example: index = 1.
• condition
This is a conditional expression that is tested every time through the loop, just as in a
while loop. It is evaluated at the beginning of every loop, and the loop is only executed
if the expression is true. Example: index <= 20.
• increment
This is an expression that is used to alter the value of the control variable. In earlier
languages, this usually meant adding or subtracting 1 from the variable. In C, it can
be almost anything. Examples: index++, index *= 20, or index /= 2.3.
For example, the following for loop prints out the integers from 1 to 10:
62 Chapter 11: Loops

int my_int;

for (my_int = 1; my_int <= 10; my_int++)

{
printf ("%d ", my_int);
printf("\n");
}

The following example prints out all prime numbers between 1 and the macro value
MAX_INT. (A prime numbers is a number that cannot be divided by any number except 1
and itself without leaving a remainder.) This program checks whether a number is a prime
by dividing it by all smaller integers up to half its size. (See Chapter 12 [Preprocessor
directives], page 67, for more information on macros.)
#include <stdio.h>

#define MAX_INT 500

#define TRUE 1
#define FALSE 0

int main ()
{
int poss_prime;

for (poss_prime = 2; poss_prime <= MAX_INT; poss_prime++)

{
if (prime(poss_prime))
{
printf ("%d ", poss_prime);
}
}
printf("\n\n");
return 0;
}

prime (int poss_prime) /* check whether poss_prime is prime */

{
int poss_factor;

for (poss_factor = 2; poss_factor <= poss_prime/2; poss_factor++)

{
if (poss_prime % poss_factor == 0)
{
return (FALSE);
}
}

return (TRUE);
}

The program should print the following sequence of integers:

2 3 5 7 11 13 17 19 23 29 31 37 41 43 47 53 59 61 67 71 73 79 83 89 97
101 103 107 109 113 127 131 137 139 149 151 157 163 167 173 179 181 191
193 197 199 211 223 227 229 233 239 241 251 257 263 269 271 277 281 283
293 307 311 313 317 331 337 347 349 353 359 367 373 379 383 389 397 401
409 419 421 431 433 439 443 449 457 461 463 467 479 487 491 499
The flexibility of for 63

11.4 The flexibility of for

As mentioned above, C’s for construct is quite versatile. You can use almost any statement
you like for its initialization, condition, and increment parts, including an empty statement.
For example, omitting the initialization and increment parts creates what is essentially a
while loop:
int my_int = 1;

for ( ; my_int <= 20; )

{
printf ("%d ", my_int);
my_int++;
}
Omitting the condition part as well produces an infinite loop, or loop that never ends:
for ( ; ; )
{
printf("Aleph Null bottles of beer on the wall...\n");
}
You can break out of an “infinite loop” with the break or return commands. (See Sec-
tion 11.5 [Terminating and speeding loops], page 64.)
Consider the following loop:
for (my_int = 2; my_int <= 1000; my_int = my_int * my_int)
{
printf ("%d ", my_int);
}
This loop begins with 2, and each time through the loop, my_int is squared.
Here’s another odd for loop:
char ch;

for (ch = ’*’; ch != ’\n’; ch = getchar())

{
/* do something */
}
This loop starts off by initializing ch with an asterisk. It checks that ch is not a linefeed
character (which it isn’t, the first time through), then reads a new value of ch with the
library function getchar and executes the code inside the curly brackets. When it detects
a line feed, the loop ends.
It is also possible to combine several increment parts in a for loop using the comma
operator ,. (See Section 18.2 [The comma operator], page 170, for more information.)
#include <stdio.h>

int main()
{
int up, down;

for (up = 0, down=10; up < down; up++, down--)

{
printf("up = %d, down= %d\n",up,down);
}

return 0;
}
64 Chapter 11: Loops

The example above will produce the following output:

up = 0, down= 10
up = 1, down= 9
up = 2, down= 8
up = 3, down= 7
up = 4, down= 6
One feature of the for loop that unnerves some programmers is that even the value of
the loop’s conditional expression can be altered from within the loop itself:
int index, number = 20;

for (index = 0; index <= number; index++)

{
if (index == 9)
{
number = 30;
}
}
In many languages, this technique is syntactically forbidden. Not so in the flexible language
C. It is rarely a good idea, however, because it can make your code confusing and hard to
maintain.

11.5 Terminating and speeding loops

C provides simple ways of terminating or speeding up any of the three loops we have
discussed, whether or not it has run its course. The three main commands to do so are
break, return, and continue.

11.5.1 Terminating loops with break

The usual statement to terminate a loop is the same statement that is used to jump out of
switch statements:
break;
If this statement is encountered within a loop, the loop will end immediately. For instance,
here is an inefficient way of assigning 12 to my_int:
for (my_int = 1; my_int <= 100; my_int++)
{
if (my_int == 12)
{
break;
}
}

printf("my_int = %d\n\n", my_int);

11.5.2 Terminating loops with return

Suppose that a program is in the middle of a loop (or some nested loops) in a complex
function, and suddenly the function finds its answer. This is where the return statement
comes in handy. The return command will jump out of any number of loops and pass
the value back to the calling function without having to finish the loops or the rest of the
function. (See Section 11.6 [Nested loops], page 65, for clarification of the idea of placing
one loop inside another.)
Example:
Nested loops 65

#include <stdio.h>

int main()
{
printf ("%d\n\n", returner(5, 10));
printf ("%d\n\n", returner(5, 5000));
return 0;
}

int returner (int foo, int bar)

{
while (foo <= 1000)
{
if (foo > bar)
{
return (foo);
}

foo++;
}

return foo;
}

The function returner contains a while loop that increments the variable foo and tests it
against a value of 1000. However, if at any point the value of foo exceeds the value of the
variable bar, the function will exit the loop, immediately returning the value of foo to the
calling function. Otherwise, when foo reaches 1000, the function will increment foo one
more time and return it to main.

Because of the values it passes to returner, the main function will first print a value of
11, then 1001. Can you see why?

11.5.3 Speeding loops with continue

Instead of terminating a loop, you might want to speed it to its next pass, perhaps to avoid
executing irrelevant statements. To do so, you should use the continue statement. When a
continue statement is encountered, the program will skip the rest of the loop’s code block
and jump straight to the start of the next pass through the loop.

Here is an example that uses the continue statement to avoid division by zero (which
causes a run-time error):
for (my_int = -10; my_int <= 10; my_int++)
{
if (my_int == 0)
{
continue;
}

printf ("%d", 20/i);

}
66 Chapter 11: Loops

11.6 Nested loops

Just as decisions can be nested, so can loops; that is, you can place loops inside other
loops. This can be useful, for example, when you are coding multidimensional arrays. (See
Chapter 14 [Arrays], page 85.)
The example below prints a square of asterisks by nesting a printf command inside an
inner loop, which is itself nested inside an outer loop.
Any kind of loop can be nested. For example, the code below could have been written
with while loops instead of for loops:
#include <stdio.h>

#define SIZE 5

int main()
{
int square_y, square_x;

printf ("\n");

for (square_y = 1; square_y <= SIZE; square_y++)

{
for (square_x = 1; square_x <= SIZE; square_x++)
{
printf("*");
}
printf ("\n");
}

printf ("\n");
return 0;
}
The output of the above code looks like this:
*****
*****
*****
*****
*****

11.7 Questions for Chapter 11

1. How many kinds of loop does C offer, and what are they?
2. When is the condition tested in each of the loops?
3. Which of the loops is always executed at least once?
4. Write a program that copies all input to output line by line.
5. Write a program to get 10 numbers from the user and add them together.
6. Rewrite the nested loops example to print a square with while loops instead of for
loops.
A few directives 67

12 Preprocessor directives

Making programming versatile.

GCC, the GNU Compiler Collection, contains a C preprocessor. A preprocessor is a

program that examines C code before it is compiled and manipulates it in various ways.
There are two main uses of a preprocessor. One is to include external files, such as header
files. The other is to define macros, which are names (possibly with arguments) that are
expanded by the preprocessor into pieces of text or C code. Macros that are expanded
into text are usually displayed to the user, but macros that are expanded into C code are
executed with the rest of the C code that surrounds them.

12.1 A few directives

All preprocessor directives, or commands, are preceded by a hash mark (‘#’). One exam-
ple you have already seen in previous chapters is the #include directive:

#include <stdio.h>

This directive tells the preprocessor to include the file ‘stdio.h’; in other words, to treat
it as though it were part of the program text.

A file to be included may itself contain #include directives, thus encompassing other
files. When this happens, the included files are said to be nested.

Here are a few other directives:

#if ... #endif

The #if directive is followed by an expression on the same line. The lines of
code between #if and #endif will be compiled only if the expression is true.
This is called conditional compilation.

#else This is part of an #if preprocessor statement and works in the same way with
#if that the regular C else does with the regular if.

#line constant filename

This causes the compiler to act as though the next line is line number constant
and is part of the file filename. Mainly used for debugging.

#error This forces the compiler to abort. Also intended for debugging.

Below is an example of conditional compilation. The following code displays ‘23’ to the
screen.
68 Chapter 12: Preprocessor directives

#include <stdio.h>

#define CHOICE 500

int my_int = 0;

#if (CHOICE == 500)

void set_my_int()
{
my_int = 23;
}
#else
void set_my_int()
{
my_int = 17;
}
#endif

int main ()
{
set_my_int();
printf("%d\n", my_int);

return 0;
}

12.2 Macros
Macros can make long, ungainly pieces of code into short words. The simplest use of macros
is to give constant values meaningful names. For example:
#define MY_PHONE 5551234
This allows the programmer to use the word MY_PHONE to mean the number 5551234.
In this case, the word is longer than the number, but it is more meaningful and makes a
program read more naturally. It can also be centralised in a header file, where it is easily
changed; this eliminates tedious search-and-replace procedures on code if the value appears
frequently in the code. It has been said, with some humorous exaggeration, that the only
values that should appear “naked” in C code instead of as macros or variables are 1 and 0.
The difference between defining a macro for 5551234 called MY_PHONE and declaring a
long integer variable called my_phone with the same value is that the variable my_phone
has the value 5551234 only provisionally; it can be incremented with the statement my_
phone++;, for example. In some sense, however, the macro MY_PHONE is that value, and
only that value — the C preprocessor simply searches through the C code before it is
compiled and replaces every instance of MY_PHONE with 5551234. Issuing the command
MY_PHONE++; is no more or less sensible than issuing the command 5551234++;.
Any piece of C code can be made into a macro, Macros are not merely constants referred
to at compile time, but are strings that are physically replaced with their values by the
preprocessor before compilation. For example:
#define SUM 1 + 2 + 3 + 4
would allow SUM to be used instead of 1 + 2 + 3 + 4. Usually, this would equal 10, so that in
the statement example1 = SUM + 10;, the variable example1 equals 20. Sometimes, though,
this macro will be evaluated differently; for instance, in the statement example2 = SUM *
Macro functions 69

10;, the variable example2 equals 46, instead of 100, as you might think. Can you figure
out why? Hint: it has to do with the order of operations.
The quotation marks in the following macro allow the string to be called by the identifier
SONG instead of typing it out over and over. Because the text ‘99 bottles of beer on the
wall...’ is enclosed by double quotation marks, it will never be interpreted as C code.
#define SONG "99 bottles of beer on the wall..."

Macros cannot define more than a single line, but they can be used anywhere except
inside strings. (Anything enclosed in string quotes is assumed to be untouchable by the
compiler.)
Some macros are defined already in the file ‘stdio.h’, for example, NULL (the value 0).
There are a few more directives for macro definition besides #define:

#undef This undefines a macro, leaving the name free.

#ifdef This is a kind of #if that is followed by a macro name. If that macro is defined
then this directive is true. #ifdef works with #else in the same way that #if
does.

#ifndef This is the opposite of #ifdef. It is also followed by a macro name. If that
name is not defined then this is true. It also works with #else.

Here is a code example using some macro definition directives from this section, and
some conditional compilation directives from the last section as well.
#include <stdio.h>

#define CHOICE 500

int my_int = 0;

#undef CHOICE
#ifdef CHOICE
void set_my_int()
{
my_int = 23;
}
#else
void set_my_int()
{
my_int = 17;
}
#endif

int main ()
{
set_my_int();
printf("%d\n", my_int);

return 0;
}

The above code example displays ‘17’ on the screen.

70 Chapter 12: Preprocessor directives

12.2.1 Macro functions

Macros can also accept parameters and return values. Macros that do so are called macro
functions. To create a macro function, simply define a macro with a parameter that has
whatever name you like, such as my_val. For example, one macro defined in the standard
libraries is abs, which returns the absolute value of its parameter. Let us define our own
version, ABS, below. (Note that we are defining it in upper case not only to avoid conflicting
with abs, but also because all macros should be defined in upper case, in the GNU coding
style. See Chapter 22 [Style], page 203.)
#define ABS(my_val) ((my_val) < 0) ? -(my_val) : (my_val)

This macro uses the ?...:... command to return a positive number no matter what value
is assigned to my_val — if my_val is defined as a positive number, the macro returns the
same number, and if my_val is defined as a negative number, the macro returns its negative
(which will be positive). (See Chapter 10 [Decisions], page 51, for more information on
the ?...:... structure. If you write ABS(-4), then the preprocessor will substitute -4 for
my_val; if you write ABS(i), then the preprocessor will substitute i for my_val, and so on.
Macros can take more than one parameter, as in the code example below.
One caveat: macros are substituted whole wherever they are used in a program: this
is potentially a huge amount of code repetition. The advantage of a macro over an actual
function, however, is speed. No time is taken up in passing control to a new function,
because control never leaves the home function; the macro just makes the function a bit
longer.
A second caveat: function calls cannot be used as macro parameters. The following code
will not work:
ABS (cos(36))

Here is an example of macro functions in use:

#include <stdio.h>

#define STRING1 "A macro definition\n"

#define STRING2 "must be all on one line!\n"
#define EXPRESSION1 1 + 2 + 3 + 4
#define EXPRESSION2 EXPRESSION1 + 10
#define ABS(x) ((x) < 0) ? -(x) : (x)
#define MAX(a,b) (a < b) ? (b) : (a)
#define BIGGEST(a,b,c) (MAX(a,b) < c) ? (c) : (MAX(a,b))

int main ()
{
printf (STRING1);
printf (STRING2);
printf ("%d\n", EXPRESSION1);
printf ("%d\n", EXPRESSION2);
printf ("%d\n", ABS(-5));
printf ("Biggest of 1, 2, and 3 is %d\n", BIGGEST(1,2,3));

return 0;
}

The output from the code example above is as follows:

Extended macro example 71

A macro definition
must be all on one line!
10
20
5
Biggest of 1, 2, and 3 is 3

12.3 Extended macro example

Here are some examples of macros taken from actual working C code, in this case the code
of GNU Emacs, the text editor of choice for many C programmers, and in fact the editor
in which this edition of the book was written.
Most of the macro examples below define various types of integer as having certain
sizes. It can be very useful when doing advanced C programming to know whether a long
integer, for instance, is 32 or 64 bits long on your system; if you select the wrong size, your
code might crash or might not even compile. In the case of Emacs, the maximum size of
certain variables (how many bits they contain) affects every aspect of its operation, even
determining how long an Emacs text file can be.
Each piece of code below is prefixed with the name of the file from which the code is
taken, and followed by a note on some interesting features of the macros defined.
• ‘emacs/src/config.h’
/* Note that lisp.h uses this in a preprocessor conditional, so it
would not work to use sizeof. That being so, we do all of them
without sizeof, for uniformity’s sake. */
#ifndef BITS_PER_INT
#define BITS_PER_INT 32
#endif

#ifndef BITS_PER_LONG
#ifdef _LP64
#define BITS_PER_LONG 64
#else
#define BITS_PER_LONG 32
#endif
#endif
In the middle of this set of macros, from ‘config.h’, the Emacs programmer used
the characters ‘/*’ and ‘*/’ to create an ordinary C comment. C comments can be
interspersed with macros freely.
The macro BITS_PER_INT is defined here to be 32 (but only if it is not already defined,
thanks to the #ifndef directive). The Emacs code will then treat integers as having
32 bits. (See Section 5.1 [Integer variables], page 17.)
The second chunk of macro code in this example checks to see whether BITS_PER_LONG
is defined. If it is not, but _LP64 is defined, it defines BITS_PER_LONG to be 64, so
that all long integers will be treated as having 64 bits. (_LP64 is a GCC macro that is
defined on 64-bit systems. It stands for “longs and pointers are 64 bits”.) If _LP64 is
not present, the code assumes it is on a 32-bit system and defines BITS_PER_LONG to
be 32.
• ‘emacs/src/lisp.h’
72 Chapter 12: Preprocessor directives

/* These are default choices for the types to use. */

#ifdef _LP64
#ifndef EMACS_INT
#define EMACS_INT long
#define BITS_PER_EMACS_INT BITS_PER_LONG
#endif
#ifndef EMACS_UINT
#define EMACS_UINT unsigned long
#endif
#else /* not _LP64 */
#ifndef EMACS_INT
#define EMACS_INT int
#define BITS_PER_EMACS_INT BITS_PER_INT
#endif
#ifndef EMACS_UINT
#define EMACS_UINT unsigned int
#endif
#endif

This set of macros, from ‘lisp.h’, again checks to see whether _LP64 is defined. If it is,
it defines EMACS_INT as long (if it is not already defined), and BITS_PER_EMACS_INT
to be the same as BITS_PER_LONG, which was defined in ‘config.h’, above. It then
defines EMACS_UINT to be an unsigned long, if it is not already defined.
If _LP64 is not defined, it is assumed we are on a 32-bit system. EMACS_INT is defined
to be an int if it is not already defined, and EMACS_UINT is defined to be an unsigned
int if it is not already defined.
Again, note that the programmer has freely interspersed a comment with the prepro-
cessor code.
• ‘emacs/src/lisp.h’
/* These values are overridden by the m- file on some machines. */
#ifndef VALBITS
#define VALBITS (BITS_PER_EMACS_INT - 4)
#endif

Here is another example from ‘lisp.h’. The macro VALBITS, which defines another
size of integer internal to Emacs, is defined as four less than BITS_PER_EMACS_INT —
that is, 60 on 64-bit systems, and 28 on 32-bit systems.
• ‘emacs/src/lisp.h’
#ifndef XINT /* Some machines need to do this differently. */
#define XINT(a) ((EMACS_INT) (((a) << (BITS_PER_EMACS_INT - VALBITS)) \
>> (BITS_PER_EMACS_INT - VALBITS)))
#endif

The interesting feature of the XINT macro above is not only that it is a function, but
that it is broken across multiple lines with the backslash character (‘\’). The GCC
preprocessor simply deletes the backslash, deletes the preceding whitespace from the
next line, and appends it where the backslash was. In this way, it is possible to treat
long, multi-line macros as though they are actually on a single line. (See Chapter 18
[Advanced operators], page 167, for more information on the the advanced operators
<< and >>.)
Questions 73

12.4 Questions
1. Define a macro called BIRTHDAY which equals the day of the month upon which your
birthday falls.
2. Write an instruction to the preprocessor to include the math library ‘math.h’.
3. A macro is always a number. True or false?
74 Chapter 12: Preprocessor directives
Header files 75

13 Libraries
Plug-in C expansions. Header files.
The core of the C language is small and simple, but special functionality is provided in
the form of external libraries of ready-made functions. Standardized libraries make C code
extremely portable, or easy to compile on many different computers.
Libraries are files of ready-compiled code that the compiler merges, or links, with a C
program during compilation. For example, there are libraries of mathematical functions,
string handling functions, and input/output functions. Indeed, most of the facilities C offers
are provided as libraries.
Some libraries are provided for you. You can also make your own, but to do so, you
will need to know how GNU builds libraries. We will discuss that later. (See Section 17.6
[Building a library], page 163.)
Most C programs include at least one library. You need to ensure both that the library
is linked to your program and that its header files are included in your program.
The standard C library, or ‘glibc’, is linked automatically with every program, but
header files are never included automatically, so you must always include them yourself.
Thus, you must always include ‘stdio.h’ in your program if you intend to use the standard
input/output features of C, even though ‘glibc’, which contains the input/output routines,
is linked automatically.
Other libraries, however, are not linked automatically. You must link them to your
program yourself. For example, to link the math library ‘libm.so’, type
gcc -o program_name program_name.c -lm
The command-line option to link ‘libm.so’ is simply ‘-lm’, without the ‘lib’ or the
‘.so’, or in the case of static libraries, ‘.a’. (See Section 13.2 [Kinds of library], page 77.)
The ‘-l’ option was created because the average GNU system already has many libraries,
and more can be added at any time. This means that sometimes two libraries provide
alternate definitions of the same function. With judicious use of the ‘-l’ option, however,
you can usually clarify to the compiler which definition of the function should be used.
Libraries specified earlier on the command line take precedence over those defined later,
and code from later libraries is only linked in if it matches a reference (function definition,
macro, global variable, etc.) that is still undefined. (See Section 17.4 [Compiling multiple
files], page 157, for more information.)
In summary, you must always do two things:
• link the library with a ‘-l’ option to gcc (a step that may be skipped in the case of
‘glibc’).
• include the library header files (a step you must always follow, even for ‘glibc’).

13.1 Header files

As mentioned above, libraries have header files that define information to be used in con-
junction with the libraries, such as functions and data types. When you include a header
file, the compiler adds the functions, data types, and other information in the header file
to the list of reserved words and commands in the language. After that, you cannot use
76 Chapter 13: Libraries

the names of functions or macros in the header file to mean anything other than what the
library specifies, in any source code file that includes the header file.
The most commonly used header file is for the standard input/output routines in ‘glibc’
and is called ‘stdio.h’. This and other header files are included with the #include com-
mand at the top of a source code file. For example,
#include "name.h"

includes a header file from the current directory (the directory in which your C source code
file appears), and
#include <name.h>

includes a file from a system directory — a standard GNU directory like ‘/usr/include’.
(The #include command is actually a preprocessor directive, or instruction to a program
used by the C compiler to simplify C code. (See Chapter 12 [Preprocessor directives],
page 67, for more information.)
Here is an example that uses the #include directive to include the standard ‘stdio.h’
header in order to print a greeting on the screen with the printf command. (The characters
‘\n’ cause printf to move the cursor to the next line.)
#include <stdio.h>

int main ()
{
printf ("C standard I/O file is included.\n");
printf ("Hello world!\n");

return 0;
}

If you save this code in a file called ‘hello.c’, you can compile this program with the
following command:
gcc -o hello hello.c

As mentioned earlier, you can use some library functions without having to link library
files explicitly, since every program is always linked with the standard C library. This
is called ‘libc’ on older operating systems such as Unix, but ‘glibc’ (“GNU libc”) on
GNU systems. The ‘glibc’ file includes standard functions for input/output, date and
time calculation, string manipulation, memory allocation, mathematics, and other language
features.
Most of the standard ‘glibc’ functions can be incorporated into your program just by
using the #include directive to include the proper header files. For example, since ‘glibc’
includes the standard input/output routines, all you need to do to be able to call printf
is put the line #include <stdio.h> at the beginning of your program, as in the example
that follows.
Note that ‘stdio.h’ is just one of the many header files you will eventually use to access
‘glibc’. The GNU C library is automatically linked with every C program, but you will
eventually need a variety of header files to access it. These header files are not included in
your code automatically — you must include them yourself!
Kinds of library 77

#include <stdio.h>
#include <math.h>

int main ()
{
double x, y;

y = sin (x);
printf ("Math library ready\n");

return 0;
}
However, programs that use a special function outside of ‘glibc’ — including mathe-
matical functions that are nominally part of ‘glibc’, such as function sin in the example
above! — must use the ‘-l’ option to gcc in order to link the appropriate libraries. If
you saved this code above in a file called ‘math.c’, you could compile it with the following
command:
gcc -o math math.c -lm
The option ‘-lm’ links in the library ‘libm.so’, which is where the mathematics routines
are actually located on a GNU system.
To learn which header files you must include in your program in order to use the features
of ‘glibc’ that interest you, consult section “Table of Contents” in The GNU C Library
Reference Manual. This document lists all the functions, data types, and so on contained
in ‘glibc’, arranged by topic and header file. (See Section 13.3 [Common library functions],
page 78, for a partial list of these header files.)
Note: Strictly speaking, you need not always use a system header file to access the
functions in a library. It is possible to write your own declarations that mimic the ones in
the standard header files. You might want to do this if the standard header files are too
large, for example. In practice, however, this rarely happens, and this technique is better
left to advanced C programmers; using the header files that came with your GNU system
is a more reliable way to access libraries.

13.2 Kinds of library

There are two kinds of library: static libraries and shared libraries. When you link to
a static library, the code for the entire library is merged with the object code for your
program. If you link to many static libraries, your executable will be enormous.
Shared libraries were developed in the late 1980s to reduce the code size of programs on
operating systems like GNU. When you link to a shared library, the library’s code is not
merged with your program’s object code. Instead, stub code is inserted into your object
code. The stub code is very small and merely calls the functions in the shared library — the
operating system does the rest. An executable created with a shared library can therefore
be far smaller than one created with a static library. Shared libraries can also reduce the
amount of memory used.
Although shared libraries seem to have every advantage over static libraries, static li-
braries are still useful. For example, sometimes you will wish to distribute an executable
to people whose computers do not have the libraries that yours does. In that case, you
might link to a static version of the libraries. This will incorporate the library functions
78 Chapter 13: Libraries

that you need into your executable, so that it will run on systems that don’t have those
libraries. (It is also sometimes easier to debug a program that is linked to static libraries
than one linked to shared libraries. See Section 23.5 [Introduction to GDB], page 213, for
more information.)
The file name for a library always starts with ‘lib’ and ends with either ‘.a’ (if it is
static) or ‘.so’ (if it is shared). For example, ‘libm.a’ is the static version of the C math
library, and ‘libm.so’ is the shared version. As explained above, you must use the ‘-l’
option with the name of a library, minus its ‘lib’ prefix and ‘.a’ or ‘.so’ suffix, to link that
library to your program (except the library ‘glibc’, which is always linked). For example,
the following shell command creates an executable program called ‘math’ from the source
code file ‘math.c’ and the library ‘libm.so’.
gcc -o math math.c -lm
The shared version of the library is always linked by default. If you want to link the static
version of the library, you must use the GCC option ‘--static’. The following example
links ‘libm.a’ instead of ‘libm.so’.
gcc -o math math.c -lm --static
Type ‘info gcc’ at your shell prompt for more information about GCC options.

13.3 Common library functions

Checking character types. Handling strings. Doing maths.
The libraries in GCC contain a repertoire of standard functions and macros. There are
many different kinds of function and macro in the libraries. Here are a few of the different
kinds available, with the header files you can use to access them:
• Character handling: ‘ctype.h’
• Mathematics: ‘math.h’
• String manipulation: ‘string.h’
You may find it useful to read the header files yourself. They are usually found in the
directories ‘/usr/include’ and its subdirectories on GNU/Linux systems. The three header
files listed above can be found in ‘/usr/include’; there is a second version of ‘ctype.h’ in
‘/usr/include/linux’.1

13.3.1 Character handling

Let’s examine some simple library functions and see how they are used. Some of the
functions that are available on GNU systems for handling individual characters are described
below. They are all macros, so the usual caveats about macro parameters apply. (See
Section 12.2.1 [Macro functions], page 70.) All of the functions below accept single variables
of type char as parameters. To use any of them, you must include the system header file
‘ctype.h’; the library used is simply glibc, which is linked automatically.

1
The version of ‘ctype.h’ in the ‘/usr/include’ directory proper is the one that comes with ‘glibc’;
the one in ‘/usr/include/linux’ is a special version associated with the Linux kernel. You can
specify the one you want with a full pathname inside double quotes (for example, #include
"/usr/include/linux/ctype.h"), or you can use the ‘-I’ option of gcc to force GCC to search a set of
directories in a specific order. See Section 17.6 [Building a library], page 163, for more information.)
Character handling 79

isalnum Returns true if and only if the parameter is alphanumeric: that is, an alphabetic
character (see isalpha) or a digit (see isdigit).
isalpha Returns true if and only if the parameter is alphabetic. An alphabetic character
is any character from ‘A’ through ‘Z’ or ‘a’ through ‘z’.
isascii Returns true if and only if the parameter is a valid ASCII character: that is, it
has an integer value in the range 0 through 127. (Remember, the char type in
C is actually a kind of integer!)
iscntrl Returns true if and only if the parameter is a control character. Control char-
acters vary from system to system, but are usually defined as characters in the
range 0 to 31.
isdigit Returns true if and only if the parameter is a digit in the range 0 through 9.
isgraph Returns true if and only if the parameter is graphic: that is, if the charac-
ter is either alphanumeric (see isalnum) or punctuation (see ispunct). All
graphical characters are valid ASCII characters, but ASCII also includes non-
graphical characters such as control characters (see iscntrl) and whitespace
(see isspace).
islower Returns true if and only if the parameter is a lower-case alphabetic character
(see isalpha).
isprint Returns true if and only if the parameter is a printable character: that is, the
character is either graphical (see isgraph) or a space character.
ispunct Returns true if and only if the parameter is a punctuation character.
isspace Returns true if and only if the parameter is a whitespace character. What
is defined as whitespace varies from system to system, but it usually includes
space characters and tab characters, and sometimes newline characters.
isupper Returns true if and only if the parameter is an upper-case alphabetic character
(see isalpha).
isxdigit Returns true if and only if the parameter is a valid hexadecimal digit: that is,
a decimal digit (see isdigit), or a letter from ‘a’ through ‘f’ or ‘A’ through ‘F’.
toascii Returns the parameter stripped of its eighth bit, so that it has an integer value
from 0 through 127 and is therefore a valid ASCII character. (See isascii.)
tolower Converts a character into its lower-case counterpart. Does not affect characters
which are already in lower case.
toupper Converts a character into its upper-case counterpart. Does not affect characters
which are already in upper case.
/********************************************************/
/* */
/* Demonstration of character utility functions */
/* */
/********************************************************/

#include <stdio.h>
#include <ctype.h>
80 Chapter 13: Libraries

#define allchars ch = 0; isascii(ch); ch++

int main () /* A criminally long main program! */

{
char ch;

printf ("\n\nVALID CHARACTERS FROM isgraph:\n\n");

for (allchars)
{
if (isgraph(ch))
{
printf ("%c ",ch);
}
}

printf ("\n\nVALID CHARACTERS FROM isalnum:\n\n");

for (allchars)
{
if (isalnum(ch))
{
printf ("%c ",ch);
}
}

printf ("\n\nVALID CHARACTERS FROM isalpha:\n\n");

for (allchars)
{
if (isalpha(ch))
{
printf ("%c ",ch);
}
}

printf ("\n\nVALID CHARACTERS FROM isupper:\n\n");

for (allchars)
{
if (isupper(ch))
{
printf ("%c ",ch);
}
}

printf ("\n\nVALID CHARACTERS FROM islower:\n\n");

for (allchars)
{
if (islower(ch))
{
printf ("%c ",ch);
}
}

printf ("\n\nVALID CHARACTERS FROM isdigit:\n\n");

for (allchars)
{
if (isdigit(ch))
{
Character handling 81

printf ("%c ",ch);

}
}

printf ("\n\nVALID CHARACTERS FROM isxdigit:\n\n");

for (allchars)
{
if (isxdigit(ch))
{
printf ("%c ",ch);
}
}

printf ("\n\nVALID CHARACTERS FROM ispunct:\n\n");

for (allchars)
{
if (ispunct(ch))
{
printf ("%c ",ch);
}
}

printf ("\n\n");

return 0;
}
The output of the above code example is as follows:
VALID CHARACTERS FROM isgraph:

! " # $ % & ’ ( ) * + , - . / 0 1 2 3 4 5 6 7 8 9 : ; < = > ? @ A B C D

E F G H I J K L M N O P Q R S T U V W X Y Z [ \ ] ^ _ ‘ a b c d e f g h
i j k l m n o p q r s t u v w x y z { | } ~

VALID CHARACTERS FROM isalnum:

0 1 2 3 4 5 6 7 8 9 A B C D E F G H I J K L M N O P Q R S T U V W X Y Z
a b c d e f g h i j k l m n o p q r s t u v w x y z

VALID CHARACTERS FROM isalpha:

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z a b c d e f g h i j
k l m n o p q r s t u v w x y z

VALID CHARACTERS FROM isupper:

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

VALID CHARACTERS FROM islower:

a b c d e f g h i j k l m n o p q r s t u v w x y z

VALID CHARACTERS FROM isdigit:

0 1 2 3 4 5 6 7 8 9

VALID CHARACTERS FROM isxdigit:

0 1 2 3 4 5 6 7 8 9 A B C D E F a b c d e f
82 Chapter 13: Libraries

VALID CHARACTERS FROM ispunct:

! " # $ % & ’ ( ) * + , - . / : ; < = > ? @ [ \ ] ^ _ ‘ { | } ~

13.4 Mathematical functions

Let us now examine some simple math library functions. (This section presupposes some
familiarity on your part with trigonometry. If you have none, you might want to skip this
section for now – but reading it won’t hurt you!)
The following mathematical functions, among others, are available on GNU systems.
Many of the functions described below are macros, so the usual caveats about macro pa-
rameters apply. (See Section 12.2.1 [Macro functions], page 70.) All of these functions
require parameters of type double or long float. Constants used must be written in
floating point form: for instance, write ‘7.0’ instead of just ‘7’.
Here is a list of some functions you can expect to find in the headers ‘math.h’, ‘tgmath.h’,
and ‘limits.h’.
abs Returns the unsigned value of the parameter in brackets. This function is a
macro; see fabs for a proper function version.
acos Returns the arccosine (or inverse cosine) of the parameter, which must lie be-
tween -1.0 and +1.0 inclusive. (The result is always in radians.)
asin Returns the arcsine (or inverse sine) of the parameter, which must lie between
-1.0 and +1.0 inclusive. (The result is always in radians.)
atan Returns the arctangent (or inverse tangent) of the parameter. (The result is
always in radians.)
atan2 This is a special function for calculating the inverse tangent of the second pa-
rameter divided by the first. atan2 will find the result more accurately than
atan will.
result = atan2 (x, y);
result = atan2 (x, 3.14);

ceil Returns the ceiling for the parameter: that is, the integer just above it. In
effect, rounds the parameter up.
cos Returns the cosine of the parameter in radians. (The parameter is also assumed
to be specified in radians.)
cosh Returns the hyperbolic cosine of the parameter.
exp Returns the exponential function of the parameter (i.e. e to the power of the
parameter).
fabs Returns the absolute or unsigned value of the parameter in brackets. This is
the version that is a proper function; see abs if you want one that is a macro.
floor Returns the floor for the parameter: that is, the integer just below it. In effect,
rounds the parameter down to the nearest integer value, i.e. truncates it.
log Returns the natural logarithm of the parameter. The parameter used must be
greater than zero, but does not have to be declared as unsigned.
Questions for Chapter 13 83

log10 Returns the base 10 logarithm of the parameter. The parameter used must be
greater than zero, but does not have to be declared as unsigned.
pow Returns the first parameter raised to the power of the second.
result = pow (x,y); /*raise x to the power y */
result = pow (x,2); /* square x */

sin Returns the sine of the parameter in radians. (The parameter is also assumed
to be specified in radians.)
sinh Returns the hyperbolic sine of the parameter. (Pronounced “shine” or “sinch”.)
sqrt Returns the positive square root of the parameter.
tan Returns the tangent of the parameter in radians. (The parameter is also as-
sumed to be specified in radians.)
tanh Returns the hyperbolic tangent of the parameter.
Here is a code example that uses a few of the math library routines listed above.
#include <stdio.h>
#include <math.h>

int main()
{
double my_pi;

my_pi = 4 * atan(1.0);

/* Print the value of pi we just calculated, to 32 digits. */

printf ("my_pi = %.32f\n", my_pi);

/* Print value of pi from math library, to 32 digits. */

printf ("M_PI = %.32f\n", M_PI);

return 0;
}
If you save the above example as ‘pi.c’, you will have to enter a command such as the
one below to compile it.
gcc pi.c -o pi -lm
When you compile and run the code example, it should print the following results:
my_pi = 3.14159265358979311599796346854419
M_PI = 3.14159265358979311599796346854419

13.5 Questions for Chapter 13

1. How do you incorporate a library file into a C program?
2. Name the most commonly used library file in C.
3. Is it possible to define new functions with the same names as standard library functions?
4. What type of data is returned from mathematical functions?
5. True or false? All mathematical calculations are performed using doubles.
6. Name five kinds of error which can occur in a mathematical function.
84 Chapter 13: Libraries
Arrays 85

14 Arrays
Rows and tables of storage.
Suppose you have a long list of numbers, but you don’t want to assign them to variables
individually. For example, you are writing a simple program for a restaurant to keep a
list of the amount each diner has on his or her tab, but you don’t want to go through the
tedium of writing a list like the following:

float alfies_tab, bettys_tab, charlies_tab ...;

alfies_tab = 88.33;
bettys_tab = 17.23;
charlies_tab = 55.55;
etc.
A list like that could run to hundreds or thousands of entries, and for each diner you’d
have to write “special-case” code referring to every diner’s data individually. No, what you
really want is a single table in which you can find the tab corresponding to a particular
diner. You can then look up the tab of the diner with dining club card number 7712 in row
number 7712 of the table much more easily.
This is why arrays were invented. Arrays are a convenient way to group many variables
under a single variable name. They are like pigeonholes, with each compartment storing a
single value. Arrays can be one-dimensional like a list, two-dimensional like a chessboard,
or three-dimensional like an apartment building — in fact, they can have any arbitrary
dimensionality, including ones humans cannot visualise easily.
An array is defined using square brackets [...]. For example: an array of three integers
called my_list would be declared thus:
int my_list[3];
This statement would cause space for three adjacent integers to be created in memory, as
in the diagram below. Notice that there is no space between the name of the array above
(my_array) and the opening square bracket ‘[’.
------------------------------------
my_list: | | | |
------------------------------------
The number in the square brackets of the declaration is referred to as the subscript of the
array, and it must be an integer greater than or equal to zero.
The three integer “pigeonholes” in the above array are called its locations, and the values
filling them are called the array’s elements. The position of an element in the array is called
its index (the plural is indices). In the following example, 5, 17, and 23 are the array’s
elements, and 0, 1, and 2 are its corresponding indices.
Notice also that although we are creating space for three integers, arrays in C are zero-
based, so the indices of the array run (0, 1, 2). If arrays in C were one-based, the indices
would run (1, 2, 3).
int my_list[3];
my_list[0] = 5;
my_list[1] = 17;
my_list[2] = 23;
86 Chapter 14: Arrays

The above example would result in an array that “looks like” the following diagram. (Of
course, an array is merely an arrangement of bytes in the computer’s memory, so it does
not look like much of anything, literally speaking.)
index: 0 1 2
------------------------------------
my_list: | 5 | 17 | 23 |
------------------------------------
Note that every element in an array must be of the same type, for example, integer. It is not
possible in C to have arrays that contain multiple data types. However, if you want an array
with multiple data types, you might instead be able to use multiple arrays of different data
types that contain the same number of elements. For example, to continue our restaurant
tab example above, one array, diner_names might contain a list of the names of the diners.
If you are looking for a particular diner, say Xavier Nougat, you might find that the index
of his name in diner_names is 7498. If you have programmed an associated floating-point
array called diner_tabs, you might look up element 7498 in that array and find that his
tab is $99.34.

14.1 Array bounds

In keeping with C’s free-wheeling, “I assume you know what you’re doing” policy, the
compiler does not complain if you try to write to elements of an array that do not exist.
For example, the code below defines an array with five elements. (Remember, C arrays are
zero-based.)
char my_array[4];
Given the line of code below, your program will happily try to write the character ‘*’ at
location 10000. Unfortunately, as may happen when writing to an uninitialized pointer,
this may crash the program, but will probably do nothing worse on a GNU system. (See
Section 9.3 [Pointers and initialization], page 46.)
my_array[10000] = ’*’;
The first and last positions in an array are called its bounds. Remember that the bounds
of an array are zero and the integer that equals the number of elements it contains, minus
one.
Although C will not warn you at compile-time when you exceed the bounds of an array,
the debugger can tell you at run-time. See Section 23.5 [Introduction to GDB], page 213,
for more information.

14.2 Arrays and for loops

When you declare an array, the computer allocates a block of memory for it, but the block
contains garbage (random values). Therefore, before using an array, you should initialise
it. It is usually a good idea to set all elements in the array to zero.
The easiest way to initialise an array is with a for loop. The following example loops
through every element in the array my_array and sets each to zero.
Remember, because arrays in C are zero-based, the indices of the array my_array in the
example below run 0 through 9, rather than 1 through 10. The effect is the same, however:
an array of ARRAY_SIZE (that is, 10) elements.
Arrays and for loops 87

#include <stdio.h>
#define ARRAY_SIZE 10

int main ()
{
int index, my_array[ARRAY_SIZE];

for (index = 0; index < ARRAY_SIZE; index++)

{
my_array[index] = 0;
printf ("my_array[%d] = %d\n", index, my_array[index]);
}
printf("\n");

return 0;
}
The output from the above example is as follows:
my_array[0] = 0
my_array[1] = 0
my_array[2] = 0
my_array[3] = 0
my_array[4] = 0
my_array[5] = 0
my_array[6] = 0
my_array[7] = 0
my_array[8] = 0
my_array[9] = 0
You can use similar code to fill the array with different values. The following code
example is nearly identical to the one above, but the line my_array[index] = index; fills
each element of the array with its own index:
#include <stdio.h>
#define ARRAY_SIZE 5

int main ()
{
int index, my_array[ARRAY_SIZE];

for (index = 0; index < ARRAY_SIZE; index++)

{
my_array[index] = index;
printf ("my_array[%d] = %d\n", index, my_array[index]);
}
printf("\n");

return 0;
}
The output is as follows:
my_array[0] = 0
my_array[1] = 1
my_array[2] = 2
my_array[3] = 3
my_array[4] = 4
Here is a human’s-eye view of the internal representation of the array (how the array "looks"
to the computer):
88 Chapter 14: Arrays

index 0 1 2 3 4
-------------------
element | 0 | 1 | 2 | 3 | 4 |
-------------------

You can use loops to do more than initialize an array. The next code example demon-
strates the use of for loops with an array to find prime numbers. The example uses a
mathematical device called the Sieve of Erastosthenes. Erastosthenes of Cyrene discovered
that one can find all prime numbers by first writing down a list of integers from 2 (the first
prime number) up to some arbitrary number, then deleting all multiples of 2 (which are
by definition not prime numbers), finding the next undeleted number after 2 (which is 3),
deleting all its multiples, finding the next undeleted number after that (5), deleting all its
multiples, and so on. When you have finished this process, all numbers that remain are
primes.
The following code example creates a Sieve of Erastosthenes for integers up to 4999,
initializes all elements with 1, then deletes all composite (non-prime) numbers by replacing
the elements that have an index equal to the composite with the macro DELETED, which
equals 0.
#include <stdio.h>

#define ARRAY_SIZE 5000

#define DELETED 0

int sieve[ARRAY_SIZE];

int main ()
{
printf ("Results of Sieve of Erastosthenes:\n\n");

fill_sieve();
delete_nonprimes();
print_primes();
}

fill_sieve ()
{
int index;

for (index = 2; index < ARRAY_SIZE; index++)

sieve[index] = 1;
}

delete_nonprimes ()
{
int index;

for (index = 2; index < ARRAY_SIZE; index++)

{
if (sieve[index] != DELETED)
delete_multiples_of_prime (index);
}
}
Multidimensional arrays 89

delete_multiples_of_prime (int prime)

{
int index, multiplier = 2;

for (index = prime * multiplier; index < ARRAY_SIZE; index = prime * multiplier++)
sieve[index] = DELETED;
}

print_primes ()
{
int index;

for (index = 2; index < ARRAY_SIZE; index++)

{
if (sieve[index] != DELETED)
printf ("%d ", index);
}
printf("\n\n");
}
Part of the output from the above program is shown below, for values up to 500. (The full
output is considerably longer.)
Results of Sieve of Erastosthenes:

2 3 5 7 11 13 17 19 23 29 31 37 41 43 47 53 59 61 67 71 73 79 83 89 97
101 103 107 109 113 127 131 137 139 149 151 157 163 167 173 179 181 191
193 197 199 211 223 227 229 233 239 241 251 257 263 269 271 277 281 283
293 307 311 313 317 331 337 347 349 353 359 367 373 379 383 389 397 401
409 419 421 431 433 439 443 449 457 461 463 467 479 487 491 499 ...

14.3 Multidimensional arrays

Suppose that you are writing a chess-playing program like GNU Chess. A chessboard is an
8-by-8 grid. What data structure would you use to represent it?
You could use an array that has a chessboard-like structure, that is, a two-dimensional
array, to store the positions of the chess pieces. Two-dimensional arrays use two indices to
pinpoint an individual element of the array. This is very similar to what is called “algebraic
notation”, already commonly used in chess circles to record games and chess problems.
In principle, there is no limit to the number of subscripts (or dimensions) an array can
have. Arrays with more than one dimension are called multidimensional arrays. Although
humans cannot easily visualize objects with more than three dimensions, representing mul-
tidimensional arrays presents no problem to computers.
In practice, however, the amount of memory in a computer tends to place limits on how
large an array can be. For example, a simple four-dimensional array of double-precision
numbers, merely twenty elements wide in each dimension, already takes up 204 ∗ 8, or
1,280,000 bytes of memory — about a megabyte. (Each element of the array takes up 8
bytes, because doubles are 64 bits wide. See Section 5.1.2 [Floating point variables], page 18,
for more information.)
You can declare an array of two dimensions as follows:
90 Chapter 14: Arrays

variable_type array_name [size1 ][size2 ]

In the above example, variable type is the name of some type of variable, such as int. Also,
size1 and size2 are the sizes of the array’s first and second dimensions, respectively. Here
is an example of defining an 8-by-8 array of integers, similar to a chessboard. Remember,
because C arrays are zero-based, the indices on each side of the chessboard array run 0
through 7, rather than 1 through 8. The effect is the same, however: a two-dimensional
array of 64 elements.
int chessboard[8][8];
To pinpoint an element in this grid, simply supply the indices in both dimensions.
Every element in this grid needs two indices to pin-point it. Normally, C program-
mers think of element 0,0 of a two-dimensional array as being the upper-left corner. The
computer, however, knows nothing of left and right, and for our purposes (attempting to
conform to international chess notation), it makes more sense to mentally “flop” the array
vertically so that element 0,0 is the lower-left corner of the board, and 7,7 the upper-right.
Thus, the first index gives the row number for the grid and the second index gives the col-
umn number. For example, 1,0 is the square directly above the lower-left corner. Suppose
that a value of 1 for an array location means a the chess king is on the space in question.
To indicate the White king’s usual position (that is, square a5 in algebraic chess notation
or 0,4 in our zero-based integer notation), you would write this:
chessboard[0][4] = 1;
Since computer memory is essentially one-dimensional, with memory locations running
straight from 0 up through the highest location in memory, a multidimensional array cannot
be stored in memory as a grid. Instead, the array is dissected and stored in rows. Consider
the following two-dimensional array.
-----------
row 0 | 1 | 2 | 3 |
-----------
row 1 | 4 | 5 | 6 |
-----------
row 2 | 7 | 8 | 9 |
-----------
Note that the numbers inside the boxes are not the actual indices of the array, which is
two-dimensional and has two indices for each element, but only arbitrary placeholders to
enable you to see which elements correspond in the following example. The row numbers
do correspond to the first index of the array, so they are numbered from 0 to 2 rather than
1 to 3.
To the computer, the array above actually “looks” like this:
------------------------------------
| 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 |
------------------------------------
| row 0 | row 1 | row 2 |
Another way of saying that arrays are stored by row is that the second index varies
fastest. A two-dimensional array is always thought of as follows:
array_name [row ][column ]
Every row stored will contain elements of many columns. The column index runs from 0
to size - 1 inside every row in the one-dimensional representation (where size is the number
Arrays and nested loops 91

of columns in the array), so the column index is changing faster than the row index, as the
one-dimensional representation of the array inside the computer is traversed.
You can represent a three-dimensional array, such as a cube, in a similar way:
variable_type array_name [size1 ][size2 ][size3 ]
Arrays do not have to be shaped like squares and cubes; you can give each dimension of
the array a different size, as follows:
int non_cube[2][6][8];
Three-dimensional arrays (and higher) are stored in the same basic way as two-dimensional
ones. They are kept in computer memory as a linear sequence of variables, and the last
index is always the one that varies fastest (then the next-to-last, and so on).

14.4 Arrays and nested loops

To initialize multidimensional arrays, you can use nested for loops. Three nested loops are
needed to initialize a three-dimensional array:
#include <stdio.h>

#define SIZE1 3
#define SIZE2 3
#define SIZE3 3

int main ()
{
int fast, faster, fastest;
int my_array[SIZE1][SIZE2][SIZE3];

for (fast = 0; fast < SIZE1; fast++)

{
for (faster = 0; faster < SIZE2; faster++)
{
for (fastest = 0; fastest < SIZE3; fastest++)
{
my_array[fast][faster][fastest] = 0;
printf("my_array[%d][%d][%d] DONE\n", fast, faster, fastest);
}
}
}
printf("\n");
}
In this example, the variables fast, faster, and fastest contain the indices of the array,
and vary fast, faster, and fastest, respectively. In the example output below, you can see
that the fastest index changes every line, while the faster index changes every three
lines, and the fast index changes only every nine lines.
my_array[0][0][0] DONE
my_array[0][0][1] DONE
my_array[0][0][2] DONE
my_array[0][1][0] DONE
my_array[0][1][1] DONE
my_array[0][1][2] DONE
my_array[0][2][0] DONE
my_array[0][2][1] DONE
my_array[0][2][2] DONE
my_array[1][0][0] DONE
92 Chapter 14: Arrays

my_array[1][0][1] DONE
my_array[1][0][2] DONE
my_array[1][1][0] DONE
my_array[1][1][1] DONE
my_array[1][1][2] DONE
my_array[1][2][0] DONE
my_array[1][2][1] DONE
my_array[1][2][2] DONE
my_array[2][0][0] DONE
my_array[2][0][1] DONE
my_array[2][0][2] DONE
my_array[2][1][0] DONE
my_array[2][1][1] DONE
my_array[2][1][2] DONE
my_array[2][2][0] DONE
my_array[2][2][1] DONE
my_array[2][2][2] DONE

Note: Although in this example we have followed the order in which indices vary inside
the computer, you do not have to do so in your own code. For example, we could have
switched the nesting of the innermost fastest and outermost fast loops, and every element
would still have been initialized. It is better, however, to be systematic about initializing
multidimensional arrays.

14.5 Initializing arrays

As mentioned above, you must initialize your arrays or they will contain garbage. There
are two main ways to do so. The first is by assigning values to array elements individually,
either as shown in the example below, or with for loops. (See Section 14.2 [Arrays and for
loops], page 86, above.)

my_array[0] = 42;
my_array[1] = 52;
my_array[2] = 23;
my_array[3] = 100;
...

The second method is more efficient and less tedious. It uses a single assignment operator
(=) and a few curly brackets ({...}).

Recall that arrays are stored by row, with the last index varying fastest. A 3 by 3 array
could be initialized in the following way:

int my_array[3][3] =
{
{10, 23, 42},
{1, 654, 0},
{40652, 22, 0}
};

Here is a small program that uses the above initialization:

Arrays as Parameters 93

#include <stdio.h>

int main()
{
int row, column;

int my_array[3][3] =
{
{10, 23, 42},
{1, 654, 0},
{40652, 22, 0}
};

for (row = 0; row <=2; row++)

{
for (column = 0; column <= 2; column++)
{
printf("%d\t", my_array[row][column]);
}
printf("\n");
}

printf("\n");

return 0;
}

The internal curly brackets are unnecessary, but they help to distinguish the rows of the
array. The following code has the same effect as the first example:
int my_array[3][3] =
{
10, 23, 42,
1, 654, 0,
40652, 22, 0
};

The same array initialization could even be written this way:

int my_array[3][3] =
{10, 23, 42, 1, 654, 0, 40652, 22, 0};

Using any of these three array initializations, the program above will print the following
text:
10 23 42
1 654 0
40652 22 0

Note 1: Be careful to place commas after every array element except the last one before a
closing curly bracket (‘}’). Be sure you also place a semicolon after the final curly bracket
of an array initializer, since here curly brackets are not delimiting a code block.
Note 2: All the expressions in an array initializer must be constants, not variables; that
is, values such as 235 and ’q’ are acceptable, depending on the type of the array, but
expressions such as the integer variable my_int are not.
Note 3: If there are not enough expressions in the array initializer to fill the array, the
remaining elements will be set to 0 if the array is static, but will be filled with garbage
otherwise.
94 Chapter 14: Arrays

14.6 Arrays as Parameters

There will be times when you want to pass an array as a parameter to a function. (For
example, you might want to pass an array of numbers to a function that will sort them.)

In the following example, notice how the array my_array in main is passed to the function
multiply as an actual parameter with the name my_array, but that the formal parameter
in the multiply function is defined as int *the_array: that is, an integer pointer. This
is the basis for much that you will hear spoken about the “equivalence of pointers and
arrays” — much that is best ignored until you have more C programming experience. The
important thing to understand is that arrays passed as parameters are considered to be
pointers by the functions receiving them. Therefore, they are always variable parameters,
which means that other functions can modify the original copy of the variable, just as
the function multiply does with the array my_array below. (See Chapter 8 [Parameters],
page 37.)

#include <stdio.h>

void multiply (int *, int);

int main()
{
int index;
int my_array[5] = {0, 1, 2, 3, 4};

multiply (my_array, 2);

for (index = 0; index < 5; index++)

printf("%d ", my_array[index]);

printf("\n\n");
return 0;
}

void multiply (int *the_array, int multiplier)

{
int index;
for (index = 0; index < 5; index++)
the_array[index] *= multiplier;
}

Even though the function multiply is declared void and therefore does not return a result,
it can still modify my_array directly, because it is a variable parameter. Therefore, the
result of the program above is as follows:

0 2 4 6 8

If you find the interchangeability of arrays and pointers as formal parameters in function
declarations to be confusing, you can always avoid the use of pointers, and declare formal
parameters to be arrays, as in the new version of the multiply function below. The result
is the same.
Questions for Chapter 14 95

void multiply (int the_array[], int multiplier)

{
int index;
for (index = 0; index < 5; index++)
the_array[index] *= multiplier;
}

14.7 Questions for Chapter 14

1. Declare an array of type double, measuring 4 by 5 elements.
2. How do you pass an array as a parameter?
3. When an array parameter is received by a function, does C allocate space for a local
variable and copy the whole array to the new location?
4. What does it mean to say that one dimension of an array “varies fastest”?
5. Which dimension of an array varies fastest, the first or the last?
96 Chapter 14: Arrays
Initializing strings 97

15 Strings

Communication using arrays.

So far we have examined variables that can contain integers, floating-point numbers, and
values that represent individual text characters. But what if you need a variable that can
contain a sequence of text characters, such as a name in a database of diners at a restaurant,
as in the examples for last chapter? That’s where strings and string variables come in.
A string value is a sequence of text characters that can become a value for a string
variable. Both a string value and a string variable can be referred to as a string, depending
on context.
In C, a string value is represented by text characters enclosed by double quotes:
"This is a string value."

A string can contain any character, including special control characters, such as the tab
character ‘\t’, the newline character ‘\n’, the “bell” character ‘\7’ (which causes the ter-
minal to beep when it is displayed), and so on.
We have been using string values since we introduced the printf command early in the
book. (See Chapter 3 [The form of a C program], page 7.) To cause your terminal to beep
twice, include the following statement in a C program:
printf("This is a string value. Beep! Beep! \7\7");

15.1 Conventions and declarations

Do not confuse strings in C with individual characters. By convention, individual characters
are enclosed in single quotes, like this: ’a’, and have the variable type char. On the other
hand, string values are enclosed in double quotes, like this: "abcdefg". String variables are
either arrays of type char or have the type “pointer to char”, that is, char *.
Conceptually, a string is an array of characters (type char). In C, string variables
can theoretically be of any length, unlike languages such as Pascal where strings hold a
maximum of 255 characters. However, the length of the string value is determined by the
position of the first null character (’/0’) in the string. Even though a string variable might
be 32,000 characters long, or longer, if the null character first appears at position 5 in the
character array, the string value is considered to be of length 5 (and contains the characters
in positions 0 through 4, in sequence). You will rarely need to consider this end marker, as
most functions in C’s string library add it or remove it automatically.

15.2 Initializing strings

Initializing string variables (or character arrays) with string values is in many ways even
easier than initializing other kinds of arrays. There are three main ways of assigning string
constants to string variables. (A string constant is a string value that was typed into the
source code, as opposed to one that is generated by the program or entered by the user.)
98 Chapter 15: Strings

#include <stdio.h>
#include <string.h>

int main()
{
/* Example 1 */
char string1[] = "A string declared as an array.\n";

/* Example 2 */
char *string2 = "A string declared as a pointer.\n";

/* Example 3 */
char string3[30];
strcpy(string3, "A string constant copied in.\n");

printf (string1);
printf (string2);
printf (string3);

return 0;
}
1. char string1[] = "A string declared as an array.\n";
This is usually the best way to declare and initialize a string. The character array is
declared explicitly. There is no size declaration for the array; just enough memory is
allocated for the string, because the compiler knows how long the string constant is.
The compiler stores the string constant in the character array and adds a null character
(‘\0’) to the end.
2. char *string2 = "A string declared as a pointer.\n";
The second of these initializations is a pointer to an array of characters. Just as in the
last example, the compiler calculates the size of the array from the string constant and
adds a null character. The compiler then assigns a pointer to the first character of the
character array to the variable string2.
Note: Most string functions will accept strings declared in either of these two ways.
Consider the printf statements at the end of the example program above — the
statements to print the variables string1 and string2 are identical.
3. char string3[30];
Declaring a string in this way is useful when you don’t know what the string variable
will contain, but have a general idea of the length of its contents (in this case, the
string can be a maximum of 30 characters long). The drawback is that you will either
have to use some kind of string function to assign the variable a value, as the next line
of code does ( strcpy(string3, "A string constant copied in.\n");), or you will
have to assign the elements of the array the hard way, character by character. (See
Section 15.4 [String library functions], page 99, for more information on the function
strcpy.)

15.3 String arrays

Suppose you want to print out a screenful of text instead of a single line. You could use one
long character array, interspersed with ‘\n’ characters where you want the lines to break,
but you might find it easier to use a string array. A string array is an array of strings, which,
String library functions 99

of course, are themselves arrays of characters; in effect, a string array is a two-dimensional

character array.
Just as there are easy methods of initializing integer arrays, float arrays, strings, and
so on, there is also an easy way of initializing string arrays. For example, here is a sample
program which prints out a menu for a full application program. That’s all it does, but you
might imagine that when the user chooses ‘3’ in the full program, the application invokes
the function calculate_bill we examined earlier. (See Chapter 8 [Parameters], page 37.)
#include <stdio.h>

char *menu[] =
{
" -------------------------------------- ",
" | ++ MENU ++ |",
" | ~~~~~~~~~~~~ |",
" | (0) Edit Preferences |",
" | (1) Print Charge Sheet |",
" | (2) Print Log Sheet |",
" | (3) Calculate Bill |",
" | (q) Quit |",
" | |",
" | |",
" | Please enter choice below. |",
" | |",
" -------------------------------------- "
};

int main()
{
int line_num;

for (line_num = 0; line_num < 13; line_num++)

{
printf ("%s\n", menu[line_num]);
}

return 0;
}

Notice that the string array menu is declared char *menu[]. This method of defining
a two-dimensional string array is a combination of methods 1 and 2 for initializing strings
from the last section. (See Section 15.2 [Initializing strings], page 97.) This is the most
convenient method; if you try to define menu with char menu[][], the compiler will return
an “unspecified bounds error”. You can get around this by declaring the second subscript
of menu explicitly (e.g. char menu[][80]), but that necessitates you know the maximum
length of the strings you are storing in the array, which is something you may not know
and that it may be tedious to find out.
The elements of menu are initialized with string constants in the same way that an integer
array, for example, is initialized with integers, separating each element with a comma. (See
Section 14.5 [Initializing arrays], page 92.)
100 Chapter 15: Strings

15.4 String library functions

The GNU C Library provides a number of very useful functions which handle strings. Here
is a list of the more common ones. To use the functions beginning with ‘ato’, you must
include the header file ‘stdlib.h’; to use the functions beginning with ‘str’, you must
include the header file ‘string.h’.
• atof Converts an ASCII string to its floating-point equivalent; for example, converts
‘-23.5’ to the value -23.5.
#include <stdio.h>
#include <stdlib.h>

int main()
{
double my_value;
char my_string[] = "+1776.23";
my_value = atof(my_string);
printf("%f\n", my_value);

return 0;
}
The output from the above code is ‘1776.230000’.
• atoi Converts an ASCII string to its integer equivalent; for example, converts ‘-23.5’
to the value -23.
int my_value;
char my_string[] = "-23.5";
my_value = atoi(my_string);
printf("%d\n", my_value);
• atol Converts an ASCII string to its long integer equivalent; for example, converts
‘+2000000000’ to the value 2000000000.
long my_value;
char my_string[] = "+2000000000";
my_value = atol(my_string);
printf("%ld\n", my_value);
• strcat Concatenates two strings: that is, joins them together into one string. Example:
char string1[50] = "Hello, ";
char string2[] = "world!\n";
strcat (string1, string2);
printf (string1);
The example above attaches the contents of string2 to the current contents of string1.
The array string1 then contains the string ‘Hello, world!\n’.
Notice that string1 was declared to be 50 characters long, more than enough to contain
the initial values of both string1 and string2. You must be careful to allocate enough
space in the string variable that will receive the concatenated data; otherwise, your
program is likely to crash. Again, on a GNU system, although your program won’t
run, nothing more drastic than an error message from the operating system is likely to
occur in such a case.
• strcmp Compares two strings and returns a value that indicates which string comes
first alphabetically. Example:
String library functions 101

int comparison;
char string1[] = "alpha";
char string2[] = "beta";

comparison = strcmp (string1, string2);

printf ("%d\n", comparison);
If the two strings are identical, strcmp returns 0. If the first string passed to strcmp
comes alphabetically before the second (that is, the first string is “less than” the second
one), strcmp returns a value less than 0. If the first string comes alphabetically after
the second one (that is, the first string is “greater than” the second one), strcmp
returns a value greater than zero. (Note that numbers come before letters in ASCII,
and upper-case letters come before lower-case ones.)
The example above prints out ‘-1’, because ‘alpha’ is alphabetically “less than” ‘beta’.
In all cases below, string1 comes alphabetically before string2, so strcmp(string1,
string2) returns a negative value.
‘string1’ string2
‘aaa’ ‘aab’
‘aaa’ ‘aaba’
‘aa’ ‘aaa’
• strcpy Copies a string into a string variable. Example:
char dest_string[50];
char source_string[] = "Are we not men?";

/* Example 1 */
strcpy (dest_string, source_string);
printf ("%s\n", dest_string);

/* Example 2 */
strcpy (dest_string, "Are we having fun yet?");
printf ("%s\n", dest_string);
The example above produces this output:
Are we not men?
Are we having fun yet?
Notes:
• The destination string in strcmp comes first, then the source string. This works
in exactly the opposite way from the GNU/Linux shell command, cp.
• You can use strcmp to copy one string variable into another (Example 1), or to
copy a string constant into a string variable (Example 2).
• Note the use of the characters ‘%s’ in the printf statements to display a string,
rather than ‘%d’ to display an integer or ‘%f’ to display a float.
• strlen Returns an integer that gives the length of a string in characters, not including
the null character at the end of the string. The following example displays the number
‘5’.
int string_length
char my_string[] = "fnord";

string_length = strlen (my_string);

printf ("%d\n", string_length);
102 Chapter 15: Strings

• strncat Works like strcat, but concatenates only a specified number of characters.
The example below displays the string ‘Hello, world! Bye’.
char string1[50] = "Hello, world! ";
char string2[] = "Bye now!";
strncat (string1, string2, 3);
printf ("%s\n", string1);
• strncmp Works like strcmp, but compares only a specified number of characters of
both strings. The example below displays ‘0’, because ‘dogberry’ and ‘dogwood’ are
identical for their first three characters.
int comparison;
char string1[] = "dogberry";
char string2[] = "dogwood";

comparison = strncmp (string1, string2, 3);

printf ("%d\n", comparison);
• strncpy Works like strcpy, but copies only a specified number of characters. The
example below displays the string ‘Are we’, because only the first six characters of
source_string are being copied into dest_string.
char dest_string[50];
char source_string[] = "Are we not men?";

strncpy (dest_string, source_string, 6);

printf ("%s\n", dest_string);
Note: As in strcmp, the destination string in strncmp comes first, then the source
string. This works in exactly the opposite way from the GNU/Linux shell command,
cp.
• strstr Tests whether a substring is present in a larger string. Returns a pointer to
the first occurrence of the substring in the larger string, or zero if the substring is not
present. (When the substring is empty, strstr returns a pointer to the first character
of the larger string.)
The example below displays ‘’foo’ is a substring of ’Got food?’.’.
char string1[] = "Got food?";
char string2[] = "foo";

if (strstr (string1, string2))

printf("’%s’ is a substring of ’%s’.\n", string2, string1);

15.5 Questions for Chapter 15

1. What are the three main ways of initializing string variables?
2. How would you declare an array of strings?
3. What information is returned by strlen?
4. What does the function strcat do? How about strncat?
5. Rewrite the Morse coder program more efficiently, using static strings. (See Section 10.6
[Example 15], page 55, for the original version.)
Input and output 103

16 Input and output

Input and output. Talking to the user. Why your printer is a file.
In order for a program to do anything useful, it usually must do some kind of input
and output, whether input from the keyboard and output to the screen, or input from and
output to the computer’s hard disk. While the C language itself does not provide much in
the way of input and output functions, the GNU C Library contains so many facilities for
input and output that a whole book could be written about them. In this chapter, we will
focus on the basics. For more information on the functions described in this chapter, and
many more, we urge you to consult hundefinedi [Table of Contents], page hundefinedi.
Most objects from which you can receive input and to which you can send output on a
GNU system are considered to be files — not only are files on your hard disk (such as object
code files, C source code files, and ordinary ASCII text files) considered to be files, but also
such peripherals as your printer, your keyboard, and your computer monitor. When you
write a C program that prompts the user for input from the keyboard, your program is
reading from, or accepting input from, the keyboard, in much the same way that it would
read a text string from a text file. Similarly, when your C program displays a text string
on the user’s monitor, it is writing to, or sending output to, the terminal, just as though
it were writing a text string to a text file. In fact, in many cases you’ll be using the very
same functions to read text from the keyboard and from text files, and to write text to the
terminal and to text files.
This curious fact will be explored later in the chapter. For now it is sufficient to say that
when C treats your computer’s peripherals as files, they are known as devices, and each
one has its own name, called a device name or pseudo-device name. On a GNU system,
the printer might be called ‘/dev/lp0’ (for “device line printer zero”) and the first floppy
drive might be called ‘/dev/fd0’ (for “device floppy drive zero”). (Why zero in both cases?
Most objects in the GNU environment are counted by starting with zero, rather than one
— just as arrays in C are zero-based.)
The advantage of treating devices as files is that it is often not necessary to know how
a particular device works, only that it is connected to the computer and can be written
to or read from. For example, C programs often get their input from the keyboard, which
C refers to with the file name ‘stdin’ (for “standard input”), and C programs often send
their output to the monitor’s text display, referred to as ‘stdout’. In some cases, ‘stdin’
and ‘stdout’ may refer to things other than the keyboard and monitor; for example, the
user may be redirecting the output from your program to a text file with the > command
in GNU/Linux. The beauty of the way the standard input/output library handles things is
that your program will work just the same.
Before you can read from or write to a file, you must first connect to it, or open it,
usually by either the fopen command, which returns its stream, or the open command,
which returns its file descriptor. You can open a file for reading, writing, or both. You can
also open a file for appending, that is, writing data after the current end of the file.
Files are made known to functions not by their file names, except in a few cases, but by
identifiers called “streams” or “file descriptors”. For example, printf uses a stream as an
identifier, not the name of the file. So does fclose:
fprintf (my_stream, "Just a little hello from fprintf.\n");
close_error = fclose (my_stream);
104 Chapter 16: Input and output

On the other hand, fopen takes a name, and returns a stream:

my_stream = fopen (my_filename, "w");
This is how you map from names to streams or file descriptors: you open the file (for
reading, writing, or both, or for appending), and the value returned from the open or fopen
function is the appropriate file descriptor or stream.
You can operate on a file either at a high level or at a low level. Operating on a file at a
high level means that you are using the file at a high level of abstraction. (See Chapter 1
[Introduction], page 1, to refresh your memory about the distinction between high and low
levels of abstraction.) Using high-level functions is usually safer and more convenient than
using low-level functions, so we will mostly concern ourselves with high-level functions in
this chapter, although we will touch on some low-level functions toward the end.
A high-level connection opened to a file is called a stream. A low-level connection to a
file is called a file descriptor. Streams and file descriptors have different data types, as we
shall see. You must pass either a stream or a file descriptor to most input/output functions,
to tell them which file they are operating on. Certain functions (usually high-level ones)
expect to be passed streams, while others (usually low-level ones) expect file descriptors.
A few functions will accept a simple filename instead of a stream or file descriptor, but
generally these are only the functions that initialize streams or file descriptors in the first
place.
You may consider it a nuisance to have to use a stream or a file descriptor to access
your file when a simple file name would seem to suffice, but these two mechanisms allow a
level of abstraction to exist between your code and your files. Remember the “black box”
analogy we explored at the beginning of the book. By using the data in files only through
streams or file descriptors, we are guaranteed the ability to write a rich variety of functions
that can exploit the behavior of these two “black box” abstractions.
Interestingly enough, although streams are considered to be for “high-level”
input/output, and file descriptors for “low-level” I/O, and GNU systems support both,
more Unix-like systems support streams than file descriptors. You can expect any
system running ISO C to support streams, but non-GNU systems may not support file
descriptors at all, or may only implement a subset of the GNU functions that operate on
file descriptors. Most of the file descriptor functions in the GNU library are included in
the POSIX.1 standard, however.
Once you have finished your input and output operations on the file, you must terminate
your connection to it. This is called closing the file. Once you have closed a file, you cannot
read from or write to it anymore until you open it again.
In summary, to use a file, a program must go through the following routine:
• Open the file for reading, writing, or both.
• Read from or write to the file as appropriate, using file-handling functions provided by
the GNU C Library.
• Close the file

16.1 High-level file routines

You can recognise most of the high-level input/output functions that operate on files because
they begin with the letter ‘f’; for example, the high-level function for opening a file called
Opening a file 105

fopen, as opposed to the low-level file-opening function open. Some of them are more
generalized versions of functions with which you may already be familiar; for example, the
function fprintf behaves like the familiar printf, but takes an additional parameter — a
stream — and sends all its output to that stream instead of simply sending its output to
‘stdout’, as printf does.

16.1.1 Opening a file

The main high-level function for opening files is fopen. When you open a file with the
fopen function, the GNU C Library creates a new stream and creates a connection between
the stream and a file. If you pass this function the name of a file that does not exist, that file
will be created. The fopen function normally returns a stream. A stream is a flow of data
from a source to a destination within a GNU system. Programs can read characters from or
write characters to a stream without knowing either the source or destination of the data,
which may be a file on disk, a device such as a terminal meant as a human interface, or
something entirely different. Streams are represented by variables of type FILE * — fopen
will return a null pointer if it fails to open the file.
The first parameter to this function is a string containing the filename of the file to open.
The filename string can be either a constant or a variable, as in the following two equivalent
examples:
FILE *my_stream;
my_stream = fopen ("foo", "r");

FILE *my_stream;
char my_filename = "foo";
my_stream2 = fopen (my_filename, "r");
The second parameter is a string containing one of the following sets of characters:
r Open the file for reading only. The file must already exist.
w Open the file for writing only. If the file already exists, its current contents are
deleted. If the file does not already exist, it is created.
r+ Open the file for reading and writing. The file must already exist. The contents
of the file are initially unchanged, but the file position is set to the beginning
of the file.
w+ Open the file for both writing and reading. If the file already exists, its current
contents are deleted. If the file does not already exist, it is created.
a Open the file for appending only. Appending to a file is the same as writing to
it, except that data is only written to the current end of the file. If the file does
not already exist, it is created.
a+ Open the file for both appending and reading. If the file exists, its contents are
unchanged until appended to. If the file does not exist, it is created. The initial
file position for reading is at the beginning of the file, but the file position for
appending is at the end of the file.
See Section 16.1.4 [File position], page 109, for more information on the concept of file
position.
106 Chapter 16: Input and output

You can also append the character ‘x’ after any of the strings in the table above. This
character causes fopen to fail rather than opening the file if the file already exists. If you
append ‘x’ to any of the arguments above, you are guaranteed not to clobber (that is,
accidentally destroy) any file you attempt to open. (Any other characters in this parameter
are ignored on a GNU system, but may be meaningful on other systems.)
The following example illustrates the proper use of fopen to open a text file for reading
(as well as highlighting the fact that you should clean up after yourself by closing files after
you are done with them). Try running it once, then running it a second time after creating
the text file ‘snazzyjazz.txt’ in the current directory with a GNU command such as touch
snazzyjazz.txt.
#include <stdio.h>

int main()
{
FILE *my_stream;

my_stream = fopen ("snazzyjazz.txt", "r");

if (my_stream == NULL)
{
printf ("File could not be opened\n");
}
else
{
printf ("File opened! Closing it now...\n");
/* Close stream; skip error-checking for brevity of example */
fclose (my_stream);
}
return 0;
}
See Section 16.1.2 [Closing a file], page 106, for more information on the function fclose.

16.1.2 Closing a file

The basic high-level function for closing files is fclose. Simply pass this function a stream,
and fopen will close it and break the connection to the corresponding file, first reading any
buffered input and writing any buffered output. If the file was closed successfully, fclose
will return 0; otherwise, it will return EOF.
It is important to check for errors when you close a stream to which you are writing.
For example, when fclose attempts to write the output remaining in its buffer, it might
generate an error because the disk is full. Following is an example of closing a file with
write access, while checking for errors.
#include <stdio.h>

int main()
{
FILE *my_stream;
char my_filename[] = "snazzyjazz.txt";
int close_error;

my_stream = fopen (my_filename, "w");

fprintf (my_stream, "Just a little hello from fprintf.\n");
Block input and output 107

close_error = fclose (my_stream);

if (close_error != 0)
{
printf ("File could not be closed.\n");
}
else
{
printf ("File closed.\n");
}

return 0;
}

16.1.3 Block input and output

You can use the two functions in this section, fread and fwrite, to read and write text in
blocks of fixed size, rather than by line or character. You can also use these two functions
to read and write blocks of binary data. This feature is useful if you want to read and
write data in the same format used by your program. For example, you can store an entire
multidimensional array of floating-point variables in a file with the fwrite command, then
read it back in directly later with the fread command, without any loss of precision caused
by converting the floats to strings for use with the fprintf function, for example. (The
main drawback to using binary files rather than formatted ASCII text files is that you
cannot easily read and edit the files you create with a text editor.)
For example, to write an array called my_array, containing object_count data objects
(such as integers), each of size object_size, to the stream my_stream, you might use the
following line of code:
fwrite (&my_array, object_size, object_count, my_stream);
To read my_array back from my_stream, you might then use the following line:
fread (&my_array, object_size, object_count, my_stream);
Here is a short table to help you remember the directions in which fwrite and fread
work.
fwrite from an array, to a file
fread from a file, to an array
The fwrite function takes four parameters. The first parameter to fwrite is a void
pointer (void *) to an array that contains the data that will be written to the file. The
second parameter is a variable of type size_t specifying the size of each object to be written,
and the third parameter, also of type size_t, specifies the number of those objects that
are to be written. The final parameter is the stream to be written to (type FILE *). If the
value returned by fopen does not match the third parameter passed (that is, the number
of objects to be written), then there was an error.
Like fwrite, the fread function takes four parameters. Its first parameter is a void
pointer to the array that will be written to. Its second parameter, of type size_t, specifies
how large each object to be read is, and its third parameter, of type size_t, specifies
how many of each object is to be read. The last parameter is simply the stream to read
from. Again, if the return value of this function does not match the third parameter, which
specifies how many object were to be read, there was an error.
108 Chapter 16: Input and output

Here is an example that creates an array and fills it with multiples of 2, prints it out,
writes the array’s data to a file with fwrite, zeroes the array and prints it out, reads the
data from the file back into the array with fread, then prints the array out again so you
can compare its data with the first set of data.
#include <stdio.h>

int main()
{
int row, column;
FILE *my_stream;
int close_error;
char my_filename[] = "my_numbers.dat";
size_t object_size = sizeof(int);
size_t object_count = 25;
size_t op_return;

int my_array[5][5] =
{
2, 4, 6, 8, 10,
12, 14, 16, 18, 20,
22, 24, 26, 28, 30,
32, 34, 36, 38, 40,
42, 44, 46, 48, 50
};
printf ("Initial values of array:\n");
for (row = 0; row <= 4; row++)
{
for (column = 0; column <=4; column++)
{
printf ("%d ", my_array[row][column]);
}
printf ("\n");
}

my_stream = fopen (my_filename, "w");

op_return = fwrite (&my_array, object_size, object_count, my_stream);
if (op_return != object_count)
{
printf ("Error writing data to file.\n");
}
else
{
printf ("Successfully wrote data to file.\n");
}

/* Close stream; skip error-checking for brevity of example */

fclose (my_stream);

printf ("Zeroing array...\n");

for (row = 0; row <= 4; row++)
{
for (column = 0; column <=4; column++)
{
my_array[row][column] = 0;
printf ("%d ", my_array[row][column]);
}
printf ("\n");
}
File position 109

printf ("Now reading data back in...\n");

my_stream = fopen (my_filename, "r");
op_return = fread (&my_array, object_size, object_count, my_stream);
if (op_return != object_count)
{
printf ("Error reading data from file.\n");
}
else
{
printf ("Successfully read data from file.\n");
}
for (row = 0; row <= 4; row++)
{
for (column = 0; column <=4; column++)
{
printf ("%d ", my_array[row][column]);
}
printf ("\n");
}

/* Close stream; skip error-checking for brevity of example */

fclose (my_stream);

return 0;
}

If all goes well, the code example above will produce the following output:
Initial values of array:
2 4 6 8 10
12 14 16 18 20
22 24 26 28 30
32 34 36 38 40
42 44 46 48 50
Successfully wrote data to file.
Zeroing array...
0 0 0 0 0
0 0 0 0 0
0 0 0 0 0
0 0 0 0 0
0 0 0 0 0
Now reading data back in...
Successfully read data from file.
2 4 6 8 10
12 14 16 18 20
22 24 26 28 30
32 34 36 38 40
42 44 46 48 50

If you attempt to view the file ‘my_numbers.dat’ produced by the program above with a
GNU command such as more numbers.dat, you will see only garbage, because the infor-
mation is stored in binary format, not readable by humans. After attempting to view this
binary file, your terminal may continue to show only garbage and you may have to reset it.
You may be able to do this with a menu option (if you are running gnome-terminal, for
example), or you may have to type reset blindly.
110 Chapter 16: Input and output

16.1.4 File position

When a person reads a book, her “location” in the book can be specified by a page number
(and even a line number or word number, at finer levels of detail). Just so, it is possible (and
often useful!) to know the location, or file position, of a stream reading from or writing to a
file. And just as we sometimes want to know which chapter a friend is currently reading in
a book, or to recommend that he flip backward or forward to an interesting passage, so it is
frequently useful to be able to change the current file position to access a more interesting
part of the file.
At the high level of the functions in this section, GNU treats all streams as streams of
characters — even binary streams like the one associated with the file ‘numbers.dat’ in the
example for fread and fwrite. (See Section 16.1.3 [Block input and output], page 107.)
This means that the file position of any stream is a simple character count — file position 0
means that we are reading or writing the first character in the file, file position 522 means
that we are reading the 523rd character, and so on. (Just as with arrays in C, file positions
are zero-based.)
Not only does the file position of a stream describe where in the file the stream is reading
or writing, but reading or writing on the stream advances the file position. During high-
level access to a file, you can change the file position at will. Any file that permits changing
the file position in an arbitrary way is called a random-access file. (Many years ago, the
people who invented computer jargon chose the word “random” to be part of the phrase
“random-access” because, from the point of view of the computer, a random-access file can
be read from or written to at any location, as if at random. Of course, programmers are
not working randomly; they decide where their programs should read and write. The term
RAM for random-access memory comes from the same source.)
The main high-level function to tell where the current file position is, is called appropri-
ately, ftell. It accepts a single parameter — a file stream — and returns a long integer
representing the file position.1 (See section “libc” in The GNU C Library Reference Manual,
for more information.)
The main function to seek a different file position is called fseek. It accepts three
parameters. The first parameter is the stream in question, the second is a long integer
offset, and the third parameter is a constant that specifies whether the offset is relative
to the beginning of the file (SEEK_SET), to the current file position (SEEK_CUR), or to the
end of the file (SEEK_END). The fseek function returns 0 if the operation was successful,
or a nonzero integer value otherwise. (A successful fseek operation also clears the end-of-
file indicator (see below), and discards the results of ungetc. See Section 16.3.5 [ungetc],
page 134.)
There is a simple macro called rewind that will take the file pointer back to the beginning
of the file. You must simply pass it the stream that you want to rewind; it does not return
a value. It is the same as calling fseek on the stream with an offset of 0 and a third
parameter of SEEK_SET, except that it resets the error indicator for the stream and, as
mentioned, there is no return value.
1
Since the file position is a long integer, the length of a file using one of these functions cannot be any
greater than the maximum value of a 32-bit long integer under GNU, plus one (since the file position is
zero-based) — that is, such a file cannot be any more than 2,147,483,648 bytes, or about two gigabytes
long. If you need to use longer files, you can use low-level file routines, which allow for longer files and
file positions through such 64-bit functions as lseek64.
End-of-file and error functions 111

An example of these functions will not be useful until we have introduced single-character
I/O. See Section 16.3.3 [getc and fgetc], page 131, if you want to read a code example that
uses the ftell, fseek, and rewind functions.

16.1.5 Stream buffering

When you write characters to a stream, they are not usually stored in the file on a character-
by-character basis as soon as they are written to the stream, but instead are accumulated in
a buffer first, then written to the file in a block when certain conditions are met. (A buffer
is an area of the computer’s memory that acts as a temporary holding area for input or
output.) Similarly, when you are reading characters from a stream, they are often buffered,
or stored in a buffer first.
It’s important to understand how buffering works, or you may find your programs behav-
ing in an unexpected way, reading and writing characters when you do not expect them to.
(You can bypass stream buffering entirely, however, by using low-level rather than high-level
file routines. See Section 16.5 [Low-level file routines], page 137, for more information.)
There are three main kinds of buffering you should know about:
• No buffering: When you write characters to an unbuffered stream, the operating system
writes them to the file as soon as possible.
• Line buffering: When you write characters to a line-buffered stream, the operating
system writes them to the file when it encounters a newline character.
• Full buffering: When you write characters to a fully-buffered stream, the operating
system writes them to the file in blocks of arbitrary size.
Most streams are fully buffered when you open them, and this is usually the best solution.
However, streams connected to interactive devices such as terminals are line-buffered when
you open them; yes, this means that ‘stdin’ and ‘stdout’ are line-buffered.
Having ‘stdin’ and ‘stdout’ be line-buffered is convenient, because most meaningful
chunks of data you write to them are terminated with a newline character. In order to
ensure that the data you read from or write to a fully-buffered stream shows up right away,
use the fflush function. In the jargon, this is called flushing the stream. Flushing moves
the characters from the buffer to the file, if they haven’t already been moved. After the
move, other functions can then work on the characters.2
To use fflush, simply pass the function the stream you want to flush. The fflush
function returns 0 if successful, or the value EOF (which is a macro defined in the GNU C
Library) if there was a write error.
Note that using fflush is not always necessary; output is flushed automatically when
you try to write and the output buffer is already full, when the stream is closed, when the
program exits, when an input operation on a stream actually reads data from the file, and
of course, when a newline is written to a line-buffered stream. (See Section 16.2.1.2 [fputs],
page 113, for a code example that uses fflush.)

2
Strictly speaking, there are multiple levels of buffering on a GNU system. Even after flushing characters
to a file, data from the file may remain in memory, unwritten to disk. On GNU systems, there is
an independently-running system program, or daemon, that periodically commits relevant data still in
memory to disk. Under GNU/Linux, this daemon is called ‘bdflush’.
112 Chapter 16: Input and output

16.1.6 End-of-file and error functions

If a file has been read to its end (that is, the current file position is the end of the file), a flag
indicating this, called the end-of-file indicator, will be set to TRUE. You can check whether
the end-of-file indicator has been set (and therefore whether the current file position is the
end of the file), with the feof function. This function takes a single argument (a stream),
and returns TRUE (a nonzero value) if the end of the file has been reached, and FALSE (zero)
otherwise.
Another flag, the error indicator, indicates whether an error has occurred during an
earlier operation on the stream. It returns TRUE if there has been an error, and FALSE
otherwise. You can check the error indicator for a stream with the ferror function. This
function takes a single argument (a stream), and returns TRUE (a nonzero value) if an error
has occured during an operation on the stream, and FALSE (zero) otherwise.
Unfortunately, ferror will not tell you what the error was, or when it occurred, only
whether there has been an error. To get a more detailed diagnosis, you can check the global
system variable errno. (See Section 16.5.1 [Usual file name errors], page 137.)
It is possible to reset the error and end-of-file indicators once they have been set for a
stream. To do so, simply pass the stream to the function clearerr; this will set both the
error and end-of-file indicators back to 0. The clearerr function does not return a value.
You should not simply reset the error flag and try a stream operation that failed a second
time. Because of buffering, you may lose or repeat data when writing, or access the wrong
part of the file when reading. Before you try a failed stream operation again, you should
seek to a known file position. (See Section 16.1.4 [File position], page 109.) However, most
errors cannot be recovered from anyway — trying the operation again will likely result in
the same error — so it is probably better to have your program report the error to the user
and exit than to write complicated error-recovery routines for stream operation.
An example of these functions will not be useful until we have introduced single-character
I/O. See Section 16.3.3 [getc and fgetc], page 131, if you want to read a code example that
uses the feof and ferror functions.

16.2 String output and input

We will now examine some high-level file functions for reading strings from and writing
strings to streams. The two string output methods we will examine (puts and fputs) are
very safe to use, but the input methods run from the antiquated and very dangerous gets to
the safer fgets, to getline and getdelim, two GNU-specific extensions to the C language
that are extremely safe to use.
It is important to use the safer and better GNU functions when you can. However, you
will probably still want to learn how to read and understand older (but still free) code that
is unsafe (perhaps to update it and make it safe), so this book describes functions like gets
despite the fact that they are unsafe.

16.2.1 Unformatted string output

The functions in this section are for output of strings to streams. They are generally very
safe to use.
Formatted string output 113

16.2.1.1 puts
The most convenient function for printing a simple message on standard outout is puts. It
is even simpler than printf, since you do not need to include a newline character — puts
does that for you.
Using puts couldn’t be simpler. Here is an example:
puts ("Hello, multiverse.");

This code example will print the string ‘Hello, multiverse.’ to standard output.
The puts function is safe and simple, but not very flexible. See Section 16.2.2 [Formatted
string output], page 114, if you want to print fancier output.

16.2.1.2 fputs
The fputs (“file put string”) function is similar to the puts function in almost every respect,
except that it accepts a second parameter, a stream to which to write the string. It does
not add a newline character, however; it only writes the characters in the string. It returns
EOF if an error occurs; otherwise it returns a non-negative integer value.
Here is a brief code example that creates a text file and uses fputs to write into it the
phrase ‘If it’s not too late... make it a cheeseburger.’, followed by a newline char-
acter. This example also demonstrates the use of the fflush function. (See Section 16.1.5
[Stream buffering], page 111, for more information on this function.)
#include <stdio.h>

int main()
{
FILE *my_stream;
char my_filename[] = "snazzyjazz.txt";
int flush_status;

my_stream = fopen (my_filename, "w");

fputs ("If it’s not too late... make it a cheeseburger.\n", my_stream);

/*
Since the stream is fully-buffered by default, not line-buffered,
it needs to be flushed periodically. We’ll flush it here for
demonstration purposes, even though we’re about to close it.
*/
flush_status = fflush (my_stream);
if (flush_status != 0)
{
puts ("Error flushing stream!");
}
else
{
puts ("Stream flushed.");
}

/* Close stream; skip error-checking for brevity of example */

fclose (my_stream);

return 0;
}
114 Chapter 16: Input and output

16.2.2 Formatted string output

The functions in this section are for formatted output of strings to streams. They are
generally quite safe to use.
Formatted output is textual output via functions such as printf or fprintf. These take
as an argument a string containing special character sequences such as ‘%d’ (which indicates
that an integer argument will follow). After this string, other arguments that correspond to
the special character sequences follow. When the functions combine these arguments, the
result is formatted textual output.
The next several sections discuss four formatted output functions. The most basic,
printf, prints to standard output. The fprintf function is a high-level routine that sends
its output to a stream, sprintf “prints” to a string, and asprintf is a safer way of printing
to a string.

16.2.2.1 printf
If you have been reading the book closely up to this point, you have seen the use of the
printf function many times. To recap, this function prints a text string to the terminal
(or, to be more precise, the text stream ‘stdout’). For example, the following line of code
prints the string ‘Hello there!’, followed by a newline character, to the console:
printf ("Hello there!\n");
You probably also remember that you can incorporate numeric constants and variables
into your strings. Consider the following code example:
printf ("I’m free! I’m free! (So what? I’m %d.)\n", 4);
The previous example is equivalent to the following one:
int age = 4;
printf ("I’m free! I’m free! (So what? I’m %d.)\n", age);
Both of the code examples above produce the following output:
I’m free! I’m free! (So what? I’m 4.)
You may recall that besides using ‘%d’ with printf to print integers, we have also used
‘%f’ on occasion to print floating-point numbers, and that on occasion we have used more
than one argument. Consider this example:
printf ("I’m free! I’m free! (So what? I’m %d.) Well, I’m %f.\n", 4, 4.5);
That example produces the following output:
I’m free! I’m free! (So what? I’m 4.) Well, I’m 4.500000.
In fact, printf is a very flexible function. The general scheme is that you provide it with
a format string or template string (such as ‘"So what? I’m %d."’), which can contain zero
or more conversion specifications, conversion specifiers, or sometimes just conversions (in
this case ‘%d’), and zero or more arguments (for example, ‘4’). Each conversion specification
is said to specify a conversion, that is, how to convert its corresponding argument into a
printable string. After the template string, you supply one argument for each conversion
specifier in the template string. The printf function then prints the template string,
including each argument as converted to a printable sub-string by its conversion specifier,
and returns an integer containing the number of characters printed, or a negative value if
there was an error.
Formatted output conversion specifiers 115

16.2.2.2 Formatted output conversion specifiers

There are many different conversion specifiers that can be used for various data types.
Conversion specifiers can become quite complex; for example, ‘%-17.7ld’ specifies that
printf should print the number left-justified (‘-’), in a field at least seventeen characters
wide (‘17’), with a minimum of seven digits (‘.7’), and that the number is a long integer
(‘l’) and should be printed in decimal notation (‘%d’).
In this section, we will examine the basics of printf and its conversion specifiers. (For
even more detail, section “Formatted Output” in The GNU C Library Reference Manual.)
A conversion specifier begins with a percent sign, and ends with one of the following
output conversion characters. The most basic conversion specifiers simply use a percent
sign and one of these characters, such as ‘%d’ to print an integer. (Note that characters in
the template string that are not part of a conversion specifier are printed as-is.)
‘c’ Print a single character.
‘d’ Print an integer as a signed decimal number.
‘e’ Print a floating-point number in exponential notation, using lower-case letters.
The exponent always contains at least two digits. Example: ‘6.02e23’.
‘E’ Same as ‘e’, but uses upper-case letters. Example: ‘6.02E23’.
‘f’ Print a floating-point number in normal, fixed-point notation.
‘i’ Same as ‘d’.
‘m’ Print the string corresponding to the specified value of the system errno vari-
able. (See Section 16.5.1 [Usual file name errors], page 137.) GNU systems
only.
‘s’ Print a string.
‘u’ Print an unsigned integer.
‘x’ Print an integer as an unsigned hexadecimal number, using lower-case letters.
‘X’ Same as ‘x’, but uses upper-case letters.
‘%’ Print a percent sign (‘%’).
In between the percent sign (‘%’) and the output conversion character, you can place
some combination of the following modifiers. (Note that the percent sign conversion (‘%%’)
doesn’t use arguments or modifiers.)
• Zero or more flag characters, from the following table:
‘-’
Left-justify the number in the field (right justification is the default). Can
also be used for string and character conversions (‘%s’ and ‘%c’).
‘+’
Always print a plus or minus sign to indicate whether the number is positive
or negative. Valid for ‘%d’, ‘%e’, ‘%E’, and ‘%i’.
116 Chapter 16: Input and output

‘Space character’
If the number does not start with a plus or minus sign, prefix it with a
space character instead. This flag is ignored if the ‘+’ flag is specified.
‘#’
For ‘%e’, ‘%E’, and ‘%f’, forces the number to include a decimal point, even
if no digits follow. For ‘%x’ and ‘%X’, prefixes ‘0x’ or ‘0X’, respectively.
‘’’
Separate the digits of the integer part of the number into groups, using
a locale-specific character. In the United States, for example, this will
usually be a comma, so that one million will be rendered ‘1,000,000’.
GNU systems only.
‘0’
Pad the field with zeroes instead of spaces; any sign or indication of base
(such as ‘0x’) will be printed before the zeroes. This flag is ignored if the
‘-’ flag or a precision is specified.
In the example given above, ‘%-17.7ld’, the flag given is ‘-’.
• An optional non-negative decimal integer specifying the minimum field width within
which the conversion will be printed. If the conversion contains fewer characters, it
will be padded with spaces (or zeroes, if the ‘0’ flag was specified). If the conversion
contains more characters, it will not be truncated, and will overflow the field. The
output will be right-justified within the field, unless the ‘-’ flag was specified. In the
example given above, ‘%-17.7ld’, the field width is ‘17’.
• For numeric conversions, an optional precision that specifies the number of digits to be
written. If it is specified, it consists of a dot character (‘.’), followed by a non-negative
decimal integer (which may be omitted, and defaults to zero if it is). In the example
given above, ‘%-17.7ld’, the precision is ‘.7’. Leading zeroes are produced if necessary.
If you don’t specify a precision, the number is printed with as many digits as necessary
(with a default of six digits after the decimal point). If you supply an argument of zero
with and explicit precision of zero, printf will not print any characters. Specifying a
precision for a string conversion (‘%s’) indicates the maximum number of characters to
write.
• An optional type modifier character from the table below. This character specifies
the data type of the argument if it is different from the default. In the example given
above, ‘%-17.7ld’, the type modifier character is ‘l’; normally, the ‘d’ output conversion
character expects a data type of int, but the ‘l’ specifies that a long int is being used
instead.
The numeric conversions usually expect an argument of either type int, unsigned
int, or double. (The ‘%c’ conversion converts its argument to unsigned char.) For
the integer conversions (‘%d’ and ‘%i’), char and short arguments are automatically
converted to type int, and for the unsigned integer conversions (‘%u’, ‘%x’, and ‘%X’),
they are converted to type unsigned int. For the floating-point conversions (‘%e’, ‘%E’,
and ‘%f’), all float arguments are converted to type double. You can use one of the
type modifiers from the table below to specify another type of argument.
Formatted output conversion specifiers 117

‘l’ Specifies that the argument is a long int (for ‘%d’ and ‘%i’), or an
unsigned long int (for ‘%u’, ‘%x’, and ‘%X’).

‘L’ Specifies that the argument is a long double for the floating-point conver-
sions (‘%e’, ‘%E’, and ‘%f’). Same as ‘ll’, for integer conversions (‘%d’ and
‘%i’).

‘ll’ Specifies that the argument is a long long int (for ‘%d’ and ‘%i’). On
systems that do not have extra-long integers, this has the same effect as
‘l’.

‘q’ Same as ‘ll’; comes from calling extra-long integers “quad ints”.

‘z’ Same as ‘Z’, but GNU only, and deprecated.

‘Z’ Specifies that the argument is of type size_t. (The size_t type is used to
specify the sizes of blocks of memory, and many functions in this chapter
use it.)

Make sure that your conversion specifiers use valid syntax; if they do not, if you do not
supply enough arguments for all conversion specifiers, or if any arguments are of the wrong
type, unpredictable results may follow. Supplying too many arguments is not a problem,
however; the extra arguments are simply ignored.
Here is a code example that shows various uses of printf.
#include <stdio.h>
#include <errno.h>

int main()
{
int my_integer = -42;
unsigned int my_ui = 23;
float my_float = 3.56;
double my_double = 424242.171717;
char my_char = ’w’;
char my_string[] = "Pardon me, may I borrow your nose?";

printf ("Integer: %d\n", my_integer);

printf ("Unsigned integer: %u\n", my_ui);

printf ("The same, as hexadecimal: %#x %#x\n", my_integer, my_ui);

printf ("Floating-point: %f\n", my_float);

printf ("Double, exponential notation: %17.11e\n", my_double);

printf ("Single character: %c\n", my_char);

printf ("String: %s\n", my_string);

errno = EACCES;
printf ("errno string (EACCES): %m\n");

return 0;
}

The code example above produces the following output on a GNU system:
118 Chapter 16: Input and output

Integer: -42
Unsigned integer: 23
The same, as hexadecimal: 0xffffffd6 0x17
Floating-point: 3.560000
Double, exponential notation: 4.24242171717e+05
Single character: w
String: Pardon me, may I borrow your nose?
errno string (EACCES): Permission denied

16.2.3 fprintf
The fprintf (“file print formatted”) command is identical to printf, except that its first
parameter is a stream to which to send output. The following code example is the same as
the one for printf, except that it sends its output to the text file ‘snazzyjazz.txt’.
#include <stdio.h>
#include <errno.h>

int main()
{
int my_integer = -42;
unsigned int my_ui = 23;
float my_float = 3.56;
double my_double = 424242.171717;
char my_char = ’w’;
char my_string[] = "Pardon me, may I borrow your nose?";

FILE *my_stream;
char my_filename[] = "snazzyjazz.txt";
my_stream = fopen (my_filename, "w");

fprintf (my_stream, "Integer: %d\n", my_integer);

fprintf (my_stream, "Unsigned integer: %u\n", my_ui);

fprintf (my_stream, "The same, as hexadecimal: %#x %#x\n", my_integer, my_ui);

fprintf (my_stream, "Floating-point: %f\n", my_float);

fprintf (my_stream, "Double, exponential notation: %17.11e\n", my_double);

fprintf (my_stream, "Single character: %c\n", my_char);

fprintf (my_stream, "String: %s\n", my_string);

errno = EACCES;
fprintf (my_stream, "errno string (EACCES): %m\n");

/* Close stream; skip error-checking for brevity of example */

fclose (my_stream);

return 0;
}

16.2.4 asprintf
The asprintf (mnemonic: “allocating string print formatted”) command is identical to
printf, except that its first parameter is a string to which to send output. It terminates
the string with a null character. It returns the number of characters stored in the string,
not including the terminating null.
sprintf 119

The asprintf function is nearly identical to the simpler sprintf, but is much safer,
because it dynamically allocates the string to which it sends output, so that the string will
never overflow. The first parameter is a pointer to a string variable, that is, it is of type
char **. The return value is the number of characters allocated to the buffer, or a negative
value if an error occurred.

The following code example prints the string ‘Being 4 is cool, but being free is
best of all.’ to the string variable my_string, then prints the string on the screen.
Notice that my_string is not initially allocated any space at all; asprintf allocates the
space itself. (See Section 16.2.1.1 [puts], page 113, for more information on the puts
function.)

#include <stdio.h>

int main()
{
char *my_string;

asprintf (&my_string, "Being %d is cool, but being free is best of all.", 4);
puts (my_string);

return 0;
}

16.2.5 Deprecated formatted string output functions

This section discusses unsafe functions for formatted string output. It actually contains
only one function, sprintf. You should never use the sprintf function; use asprintf
instead.

16.2.5.1 sprintf
The sprintf (“string print formatted”) command is similar to asprintf, except that it is
much less safe. Its first parameter is a string to which to send output. It terminates the
string with a null character. It returns the number of characters stored in the string, not
including the terminating null.

This function will behave unpredictably if the string to which it is printing overlaps any
of its arguments. It is dangerous because the characters output to the string may overflow
it. This problem cannot be solved with the field width modifier to the conversion specifier,
because only the minimum field width can be specified with it. To avoid this problem,
it is better to use asprintf, but there is a lot of C code that still uses sprintf, so it is
important to know about it. (See Section 16.2.4 [asprintf], page 118.)

The following code example prints the string ‘Being 4 is cool, but being free is
best of all.’ to the string variable my_string then prints the string on the screen.
Notice that my_string has been allocated 100 bytes of space, enough to contain the
characters output to it. (See Section 16.2.1.1 [puts], page 113, for more information on the
puts function.)
120 Chapter 16: Input and output

#include <stdio.h>

int main()
{
char my_string[100];

sprintf (my_string, "Being %d is cool, but being free is best of all.", 4);
puts (my_string);

return 0;
}

16.2.6 String input

The functions in this section are for input of strings from streams. They are generally very
safe to use.

16.2.6.1 getline
The getline function is the preferred method for reading lines of text from a stream,
including standard input. The other standard functions, including gets, fgets, and scanf,
are too unreliable. (Doubtless, in some programs you will see code that uses these unreliable
functions, and at times you will come across compilers that cannot handle the safer getline
function. As a professional, you should avoid unreliable functions and any compiler that
requires you to be unsafe.)
The getline function reads an entire line from a stream, up to and including the next
newline character. It takes three parameters. The first is a pointer to a block allocated
with malloc or calloc. (These two functions allocate computer memory for the program
when it is run. See Section 20.2 [Memory allocation], page 189, for more information.) This
parameter is of type char **; it will contain the line read by getline when it returns. The
second parameter is a pointer to a variable of type size_t; this parameter specifies the size
in bytes of the block of memory pointed to by the first parameter. The third parameter is
simply the stream from which to read the line.
The pointer to the block of memory allocated for getline is merely a suggestion. The
getline function will automatically enlarge the block of memory as needed, via the realloc
function, so there is never a shortage of space — one reason why getline is so safe. Not
only that, but getline will also tell you the new size of the block by the value returned in
the second parameter.
If an error occurs, such as end of file being reached without reading any bytes, getline
returns -1. Otherwise, the first parameter will contain a pointer to the string containing the
line that was read, and getline returns the number of characters read (up to and including
the newline, but not the final null character). The return value is of type ssize_t.
Although the second parameter is of type pointer to string (char **), you cannot treat
it as an ordinary string, since it may contain null characters before the final null character
marking the end of the line. The return value enables you to distinguish null characters
that getline read as part of the line, by specifying the size of the line. Any characters in
the block up to the number of bytes specified by the return value are part of the line; any
characters after that number of bytes are not.
Here is a short code example that demonstrates how to use getline to read a line of
text from the keyboard safely. Try typing more than 100 characters. Notice that getline
gets 121

can safely handle your line of input, no matter how long it is. Also note that the puts
command used to display the line of text read will be inadequate if the line contains any
null characters, since it will stop displaying text at the first null, but that since it is difficult
to enter null characters from the keyboard, this is generally not a consideration.
#include <stdio.h>

int main()
{
int bytes_read;
int nbytes = 100;
char *my_string;

puts ("Please enter a line of text.");

/* These 2 lines are the heart of the program. */

my_string = (char *) malloc (nbytes + 1);
bytes_read = getline (&my_string, &nbytes, stdin);

if (bytes_read == -1)
{
puts ("ERROR!");
}
else
{
puts ("You typed:");
puts (my_string);
}

return 0;
}

16.2.6.2 getdelim
The getdelim function is a more general form of the getline function; whereas getline
stops reading input at the first newline character it encounters, the getdelim function
enables you to specify other delimiter characters than newline. In fact, getline simply
calls getdelim and specifies that the delimiter character is a newline.
The syntax for getdelim is nearly the same as that of getline, except that the third
parameter specifies the delimiter character, and the fourth parameter is the stream from
which to read. You can exactly duplicate the getline example in the last section with
getdelim, by replacing the line
bytes_read = getline (&my_string, &nbytes, stdin);

with the line

bytes_read = getdelim (&my_string, &nbytes, ’\n’, stdin);

16.2.7 Deprecated string input functions

The functions in this section are for input of strings from streams, but they are generally
dangerous and should only be called when there is no alternative. They are included here
because you may come across code imported from a non-GNU system that uses these unsafe
functions.
122 Chapter 16: Input and output

16.2.7.1 gets
If you want to read a string from standard input, you can use the gets function, the name
of which stands for “get string”. However, this function is deprecated — that means it is
obsolete and it is strongly suggested you do not use it — because it is dangerous. It is
dangerous because it provides no protection against overflowing the string into which it is
saving data. Programs that use gets can actually be a security problem on your computer.
Since it is sometimes used in older code (which is why the GNU C Library still provides
it), we will examine it briefly; nevertheless, you should always use the function getline
instead. (See Section 16.2.6.1 [getline], page 120.)
The gets function takes one parameter, the string in which to store the data read. It
reads characters from standard input up to the next newline character (that is, when the
user presses hRETURNi), discards the newline character, and copies the rest into the string
passed to it. If there was no error, it returns the same string (as a return value, which may
be discarded); otherwise, if there was an error, it returns a null pointer.
Here is a short code example that uses gets:
#include <stdio.h>

int main()
{
char my_string[500];
printf("Type something.\n");
gets(my_string);
printf ("You typed: %s\n", my_string);

return 0;
}
If you attempt to compile the example above, it will compile and will run properly, but
GCC will warn you against the use of a deprecated function, as follows:
/tmp/ccPW3krf.o: In function ‘main’:
/tmp/ccPW3krf.o(.text+0x24): the ‘gets’ function
is dangerous and should not be used.
Remember! Never use this function in your own code. Always use getline instead.

16.2.7.2 fgets
The fgets (“file get string”) function is similar to the gets function. This function is
deprecated — that means it is obsolete and it is strongly suggested you do not use it —
because it is dangerous. It is dangerous because if the input data contains a null character,
you can’t tell. Don’t use fgets unless you know the data cannot contain a null. Don’t use
it to read files edited by the user because, if the user inserts a null character, you should
either handle it properly or print a clear error message. Always use getline or getdelim
instead of fgets if you can.
Rather than reading a string from standard input, as gets does, fgets reads it from
a specified stream, up to and including a newline character. It stores the string in the
string variable passed to it, adding a null character to terminate the string. This function
takes three parameters: the first is the string into which to read data, the second is the
maximum number of characters to read. (You must supply at least this many characters of
space in the string, or your program will probably crash, but at least the fgets function
protects against overflowing the string and creating a security hazard, unlike gets.) The
Formatted string input 123

third parameter is the stream from which to read. The number of characters that fgets
reads is actually one less than than number specified; it stores the null character in the
extra character space.
If there is no error, fgets returns the string read as a return value, which may be
discarded. Otherwise, for example if the stream is already at end of file, it returns a null
pointer.
Unfortunately, like the gets function, fgets is deprecated, in this case because when
fgets cannot tell whether a null character is included in the string it reads. If a null
character is read by fgets, it will be stored in the string along with the rest of the characters
read. Since a null character terminates a string in C, C will then consider your string to
end prematurely, right before the first null character. Only use fgets if you are certain the
data read cannot contain a null; otherwise, use getline.
Here is a code example that uses fgets. It will create a text file containing the string
‘Hidee ho!’ plus a newline, read it back with fgets, and print it on standard output. Notice
that although 100 characters are allocated for the string my_string, and requested to be
read in the fgets call, there are not that many characters in the file. The fgets function
only reads the string up to the newline character; the important thing is to allocate enough
space in the string variable to contain the string to be read.
#include <stdio.h>

int main()
{
int input_character;
FILE *my_stream;
char my_filename[] = "snazzyjazz.txt";
char my_string[100];

my_stream = fopen (my_filename, "w");

fprintf (my_stream, "Hidee ho!\n");

/* Close stream; skip error-checking for brevity of example */

fclose (my_stream);

my_stream = fopen (my_filename, "r");

fgets (my_string, 100, my_stream);

/* Close stream; skip error-checking for brevity of example */

fclose (my_stream);

printf ("%s", my_string);

return 0;
}

16.2.8 Formatted string input

The formatted string input functions are the opposite of the formatted string output func-
tions. Unlike printf and similar functions, which generate formatted output, scanf and
its friends parse formatted input. Like the opposite functions, each accepts, as a parameter,
a template string that contains conversion specifiers. In the case of scanf and related func-
tions, however, the conversion specifiers are meant to match patterns in an input string,
124 Chapter 16: Input and output

such as integers, floating point numbers, and character sequences, and store the values read
in variables.

16.2.8.1 sscanf
The sscanf function accepts a string from which to read input, then, in a manner similar to
printf and related functions, it accepts a template string and a series of related arguments.
It tries to match the template string to the string from which it is reading input, using
conversion specifier like those of printf.
The sscanf function is just like the deprecated parent scanf function, except that the
first argument of sscanf specifies a string from which to read, whereas scanf can only read
from standard input. Reaching the end of the string is treated as an end-of-file condition.
Here is an example of sscanf in action:
sscanf (input_string, "%as %as %as", &str_arg1, &str_arg2, &str_arg3);

If the string sscanf is scanning overlaps with any of the arguments, unexpected results
will follow, as in the following example. Don’t do this!
sscanf (input_string, "%as", &input_string);

Here is a good code example that parses input from the user with sscanf. It prompts
the user to enter three integers separated by whitespace, then reads an arbitrarily long line
of text from the user with getline. It then checks whether exactly three arguments were
assigned by sscanf. If the line read does not contain the data requested (for example, if it
contains a floating-point number or any alphabetic characters), the program prints an error
message and prompts the user for three integers again. When the program finally receives
exactly the data it was looking for from the user, it prints out a message acknowledging the
input, and then prints the three integers.
It is this flexibility of input and great ease of recovery from errors that makes the
getline/sscanf combination so vastly superior to scanf alone. Simply put, you should
never use scanf where you can use this combination instead.
#include <stdio.h>

int main()
{
int nbytes = 100;
char *my_string;
int int1, int2, int3;
int args_assigned;

args_assigned = 0;

while (args_assigned != 3)
{
puts ("Please enter three integers separated by whitespace.");
my_string = (char *) malloc (nbytes + 1);
getline (&my_string, &nbytes, stdin);
args_assigned = sscanf (my_string, "%d %d %d", &int1, &int2, &int3);
if (args_assigned != 3)
puts ("\nInput invalid!");
}
Formatted input conversion specifiers 125

printf ("\nThanks!\n%d\n%d\n%d\n", int1, int2, int3);

return 0;
}
Template strings for sscanf and related functions are somewhat more free-form than
those for printf. For example, most conversion specifiers ignore any preceding whitespace.
Further, you cannot specify a precision for sscanf conversion specifiers, as you can for those
of printf.
Another important difference between sscanf and printf is that the arguments to
sscanf must be pointers; this allows sscanf to return values in the variables they point
to. If you forget to pass pointers to sscanf, you may receive some strange errors, and it
is easy to forget to do so; therefore, this is one of the first things you should check if code
containing a call to sscanf begins to go awry.
A sscanf template string can contain any number of any number of whitespace char-
acters, any number of ordinary, non-whitespace characters, and any number of conversion
specifiers starting with ‘%’. A whitespace character in the template string matches zero or
more whitespace characters in the input string. Ordinary, non-whitespace characters must
correspond exactly in the template string and the input stream; otherwise, a matching error
occurs. Thus, the template string ‘" foo "’ matches ‘"foo"’ and ‘" foo "’, but not ‘" food
"’.
If you create an input conversion specifier with invalid syntax, or if you don’t supply
enough arguments for all the conversion specifiers in the template string, your code may do
unexpected things, so be careful. Extra arguments, however, are simply ignored.
Conversion specifiers start with a percent sign (‘%’) and terminate with a character from
the following table:

16.2.8.2 Formatted input conversion specifiers

‘c’ Matches a fixed number of characters. If you specify a maximum field width (see
below), that is how many characters will be matched; otherwise, ‘%c’ matches
one character. This conversion does not append a null character to the end
of the text it reads, as does the ‘%s’ conversion. It also does not skip white-
space characters, but reads precisely the number of characters it was told to,
or generates a matching error if it cannot.
‘d’ Matches an optionally signed decimal integer, containing the following sequence:
1. An optional plus or minus sign (‘+’ or ‘-’).
2. One or more decimal digits.
Note that ‘%d’ and ‘%i’ are not synonymous for scanf, as they are for printf.
‘e’ Matches an optionally signed floating-point number, containing the following
sequence:
1. An optional plus or minus sign (‘+’ or ‘-’).
2. A floating-point number in decimal or hexadecimal format.
• The decimal format is a sequence of one or more decimal digits, op-
tionally containing a decimal point character (usually ‘.’), followed
126 Chapter 16: Input and output

by an optional exponent part, consisting of a character ‘e’ or ‘E’, an

optional plus or minus sign, and a sequence of decimal digits.
• The hexadecimal format is a ‘0x’ or ‘0X’, followed by a sequence of
one or more hexadecimal digits, optionally containing a decimal point
character, followed by an optional binary-exponent part, consisting of
a character ‘p’ or ‘P’, an optional plus or minus sign, and a sequence
of digits.
‘E’ Same as ‘e’.
‘f’ Same as ‘e’.
‘g’ Same as ‘e’.
‘G’ Same as ‘e’.
‘i’ Matches an optionally signed integer, containing the following sequence:
1. An optional plus or minus sign (‘+’ or ‘-’).
2. A string of characters representing an unsigned integer.
• If the string begins with ‘0x’ or ‘0X’, the number is assumed to be in
hexadecimal format, and the rest of the string must contain hexadec-
imal digits.
• Otherwise, if the string begins with ‘0’, the number is assumed to be
in octal format (base eight), and the rest of the string must contain
octal digits.
• Otherwise, the number is assumed to be in decimal format, and the
rest of the string must contain decimal digits.
Note that ‘%d’ and ‘%i’ are not synonymous for scanf, as they are for printf.
You can print integers in this syntax with printf by using the ‘#’ flag char-
acter with the ‘%x’ or ‘%d’ output conversions. (See Section 16.2.2.1 [printf],
page 114.)
‘s’ Matches a string of non-whitespace characters. It skips initial whitespace, but
stops when it meets more whitespace after it has read something. It stores a
null character at the end of the text that it reads, to mark the end of the string.
(See Section 16.2.9.2 [String overflows with scanf], page 128, for a warning
about using this conversion.)
‘x’ Matches an unsigned integer in hexadecimal format. The string matched must
begin with ‘0x’ or ‘0X’, and the rest of the string must contain hexadecimal
digits.
‘X’ Same as ‘x’.
‘[’ Matches a string containing an arbitrary set of characters. For example,
‘%12[0123456789]’ means to read a string with a maximum field width of
12, containing characters from the set ‘0123456789’ — in other words, twelve
decimal digits. An embedded ‘-’ character means a range of characters; thus
‘%12[0-9]’ means the same thing as the last example. Preceding the characters
in the square brackets with a caret (‘^’) means to read a string not containing
Formatted input conversion specifiers 127

the characters listed. Thus, ‘%12[^0-9]’ means to read a twelve-character

string not containing any decimal digit. (See Section 16.2.9.2 [String overflows
with scanf], page 128, for a warning about using this conversion.)
‘%’ Matches a percent sign. Does not correspond to an argument, and does not
permit flags, field width, or type modifier to be specified (see below).
In between the percent sign (‘%’) and the input conversion character, you can place some
combination of the following modifiers, in sequence. (Note that the percent sign conversion
(‘%%’) doesn’t use arguments or modifiers.)
• An optional ‘*’ flag. This flag specifies that a match should be made between the
conversion specifier and an item in the input stream, but that the value should not
then be assigned to an argument.
• An optional ‘a’ flag, valid with string conversions only. This is a GNU extension to
scanf that requests allocation of a buffer long enough to safely store the string that
was read. (See Section 16.2.9.2 [String overflows with scanf], page 128, for information
on how to use this flag.)
• An optional ‘’’ flag. This flag specifies that the number read will be grouped according
to the rules currently specified on your system. For example, in the United States, this
usually means that ‘1,000’ will be read as one thousand.
• An optional decimal integer that specifies the maximum field width. The scanf func-
tion will stop reading characters from the input stream either when this maximum is
reached, or when a non-matching character is read, whichever comes first. Discarded
initial whitespace does not count toward this width; neither does the null character
stored by string input conversions to mark the end of the string.
• An optional type modifier character from the following table. (The default type of the
corresponding argument is int * for the ‘%d’ and ‘%i’ conversions, unsigned int * for
‘%x’ and ‘%X’, and float * for ‘%e’ and its synonyms. You can use these type modifiers
to specify otherwise.)
‘h’ Specifies that the argument to which the value read should be assigned is
of type short int * or unsigned short int *. Valid for the ‘%d’ and ‘%i’
conversions.
‘l’ For the ‘%d’ and ‘%i’ conversions, specifies that the argument to which the
value read should be assigned is of type long int * or unsigned long int
*. For the ‘%e’ conversion and its synonyms, specifies that the argument is
of type double *.
‘L’ For the ‘%d’ and ‘%i’ conversions, specifies that the argument to which the
value read should be assigned is of type long long int * or unsigned long
long int *. On systems that do not have extra-long integers, this has the
same effect as ‘l’.
For the ‘%e’ conversion and its synonyms, specifies that the argument is of
type long double *.
‘ll’ Same as ‘L’, for the ‘%d’ and ‘%i’ conversions.
‘q’ Same as ‘L’, for the ‘%d’ and ‘%i’ conversions.
128 Chapter 16: Input and output

‘z’ Specifies that the argument to which the value read should be assigned is
of type size_t. (The size_t type is used to specify the sizes of blocks of
memory, and many functions in this chapter use it.) Valid for the ‘%d’ and
‘%i’ conversions.

16.2.9 Deprecated formatted string input functions

These formatted string input functions are generally dangerous and should only be used
when there is no alternative. However, because you may encounter them when importing
older code or code from non-GNU systems, and because the scanf function is in a sense
the parent of the safe sscanf function, it is important that you know about them.

16.2.9.1 scanf
The first of the functions we will examine is scanf (“scan formatted”). The scanf function
is considered dangerous for a number of reasons. First, if used improperly, it can cause
your program to crash by reading character strings that overflow the string variables meant
to contain them, just like gets. (See Section 16.2.7.1 [gets], page 121.) Second, scanf
can hang if it encounters unexpected non-numeric input while reading a line from standard
input. Finally, it is difficult to recover from errors when the scanf template string does not
match the input exactly.
If you are going to read input from the keyboard, it is far better to read it with getline
and parse the resulting string with sscanf (“string scan formatted”) than to use scanf
directly. However, since sscanf uses nearly the same syntax as sscanf, as does the related
fscanf, and since scanf is a standard C function, it is important to learn about it.
If scanf cannot match the template string to the input string, it will return immediately
— and it will leave the first non-matching character as the next character to read from the
stream. This is called a matching error, and is the main reason scanf tends to hang
when reading input from the keyboard; a second call to scanf will almost certainly choke,
since the file position indicator of the stream is not pointing where scanf will expect it to.
Normally, scanf returns the number of assignments made to the arguments it was passed,
so check the return value to see if scanf found all the items you expected.

16.2.9.2 String overflows with scanf

If you use the ‘%s’ and ‘%[’ conversions improperly, then the number of characters read is
limited only by where the next whitespace character appears. This almost cetainly means
that invalid input could make your program crash, because input too long would overflow
whatever buffer you have provided for it. No matter how long your buffer is, a user could
always supply input that is longer. A well-written program reports invalid input with a
comprehensible error message, not with a crash.
Fortunately, it is possible to avoid scanf buffer overflow by either specifying a field width
or using the ‘a’ flag.
When you specify a field width, you need to provide a buffer (using malloc or a similar
function) of type char *. (See Section 20.2 [Memory allocation], page 189, for more infor-
mation on malloc.) You need to make sure that the field width you specify does not exceed
the number of bytes allocated to your buffer.
On the other hand, you do not need to allocate a buffer if you specify the ‘a’ flag character
— scanf will do it for you. Simply pass scanf an pointer to an unallocated variable of
fscanf 129

type char *, and scanf will allocate however large a buffer the string requires, and return
the result in your argument. This is a GNU-only extension to scanf functionality.
Here is a code example that shows first how to safely read a string of fixed maximum
length by allocating a buffer and specifying a field width, then how to safely read a string
of any length by using the ‘a’ flag.
#include <stdio.h>

int main()
{
int bytes_read;
int nbytes = 100;
char *string1, *string2;

string1 = (char *) malloc (25);

puts ("Please enter a string of 20 characters or fewer.");

scanf ("%20s", string1);
printf ("\nYou typed the following string:\n%s\n\n", string1);

puts ("Now enter a string of any length.");

scanf ("%as", &string2);
printf ("\nYou typed the following string:\n%s\n", string2);

return 0;
}

There are a couple of things to notice about this example program. First, notice that
the second argument passed to the first scanf call is string1, not &string1. The scanf
function requires pointers as the arguments corresponding to its conversions, but a string
variable is already a pointer (of type char *), so you do not need the extra layer of indirection
here. However, you do need it for the second call to scanf. We passed it an argument of
&string2 rather than string2, because we are using the ‘a’ flag, which allocates a string
variable big enough to contain the characters it read, then returns a pointer to it.
The second thing to notice is what happens if you type a string of more than 20 characters
at the first prompt. The first scanf call will only read the first 20 characters, then the
second scanf call will gobble up all the remaining characters without even waiting for a
response to the second prompt. This is because scanf does not read a line at a time,
the way the getline function does. Instead, it immediately matches attempts to match
its template string to whatever characters are in the stdin stream. The second scanf call
matches all remaining characters from the overly-long string, stopping at the first whitespace
character. Thus, if you type ‘12345678901234567890xxxxx’ in response to the first prompt,
the program will immediately print the following text without pausing:
You typed the following string:
12345678901234567890

Now enter a string of any length.

You typed the following string:

xxxxx

(See Section 16.2.8.1 [sscanf], page 124, for a better example of how to parse input from
the user.)
130 Chapter 16: Input and output

16.2.10 fscanf
The fscanf function is just like the scanf function, except that the first argument of fscanf
specifies a stream from which to read, whereas scanf can only read from standard input.
Here is a code example that generates a text file containing five numbers with fprintf,
then reads them back in with fscanf. Note the use of the ‘#’ flags in the ‘%#d’ conversions
in the fprintf call; this is a good way to generate data in a format that scanf and related
functions can easily read with the ‘%i’ input conversion.
#include <stdio.h>
#include <errno.h>

int main()
{
float f1, f2;
int i1, i2;
FILE *my_stream;
char my_filename[] = "snazzyjazz.txt";

my_stream = fopen (my_filename, "w");

fprintf (my_stream, "%f %f %#d %#d", 23.5, -12e6, 100, 5);

/* Close stream; skip error-checking for brevity of example */

fclose (my_stream);

my_stream = fopen (my_filename, "r");

fscanf (my_stream, "%f %f %i %i", &f1, &f2, &i1, &i2);

/* Close stream; skip error-checking for brevity of example */

fclose (my_stream);

printf ("Float 1 = %f\n", f1);

printf ("Float 2 = %f\n", f2);
printf ("Integer 1 = %d\n", i1);
printf ("Integer 2 = %d\n", i2);

return 0;
}
This code example prints the following output on the screen:
Float 1 = 23.500000
Float 2 = -12000000.000000
Integer 1 = 100
Integer 2 = 5
If you examine the text file ‘snazzyjazz.txt’, you will see it contains the following text:
23.500000 -12000000.000000 100 5

16.3 Single-character input and output

This section covers the use of several functions for the input and output of single characters
from standard input and output or files.

16.3.1 getchar
If you want to read a single character from standard input, you can use the getchar
function. This function takes no parameters, but reads the next character from ‘stdin’ as
getc and fgetc 131

an unsigned char, and returns its value, converted to an integer. Here is a short program
that uses getchar:
#include <stdio.h>

int main()
{
int input_character;

printf("Hit any key, then hit RETURN.\n");

input_character = getchar();
printf ("The key you hit was ’%c’.\n", input_character);
printf ("Bye!\n");

return 0;
}
Note that because stdin is line-buffered, getchar will not return a value until you hit
the hRETURNi key. However, getchar still only reads one character from stdin, so if you
type ‘hellohellohello’ at the prompt, the program above will still only get once character.
It will print the following line, and then terminate:
The key you hit was ’h’.
Bye!

16.3.2 putchar
If you want to print a single character on standard output, you can use the putchar function.
It takes a single integer parameter containing a character (the argument can be a single-
quoted text character, as in the example below), and sends the character to stdout. If a
write error occurs, putchar returns EOF; otherwise, it returns the integer it was passed.
This can simply be disregarded, as in the example below.
Here is a short code example that makes use of putchar. It prints an ‘X’, a space, and
then a line of ten exclamation marks (‘!!!!!!!!!!’) on the screen, then outputs a newline
so that the next shell prompt will not occur on the same line. Notice the use of the for
loop; by this means, putchar can be used not just for one character, but multiple times.
#include <stdio.h>

int main()
{
int i;

putchar (’X’);
putchar (’ ’);
for (i=1; i<=10; i++)
{
putchar (’!’);
}
putchar (’\n’);

return 0;
}

16.3.3 getc and fgetc

If you want to read a single character from a stream other than stdin, you can use the getc
function. This function is very similar to getchar, but accepts an argument that specifies
132 Chapter 16: Input and output

the stream from which to read. It reads the next character from the specified stream as an
unsigned char, and returns its value, converted to an integer. If a read error occurs or the
end of the file is reached, getc returns EOF instead.
Here is a code example that makes use of getc. This code example creates a text
file called ‘snazzyjazz.txt’ with fopen, writes the alphabet in upper-case letters plus a
newline to it with fprintf, reads the file position with ftell, and gets the character there
with getc. It then seeks position 25 with fseek and repeats the process, attempts to read
past the end of the file and reports end-of-file status with feof, and generates an error by
attempting to write to a read-only stream. It then reports the error status with ferror,
returns to the start of the file with rewind and prints the first character, and finally attempts
to close the file and prints a status message indicating whether it could do so.
See Section 16.1.4 [File position], page 109, for information on ftell, fseek, and rewind.
See Section 16.1.6 [End-of-file and error functions], page 112, for more information on feof
and ferror.
#include <stdio.h>

int main()
{
int input_char;
FILE *my_stream;
char my_filename[] = "snazzyjazz.txt";
long position;
int eof_status, error_status, close_error;

my_stream = fopen (my_filename, "w");

fprintf (my_stream, "ABCDEFGHIJKLMNOPQRSTUVWXYZ");

/* Close stream; skip error-checking for brevity of example */

fclose (my_stream);

printf ("Opening file...\n");

my_stream = fopen (my_filename, "r");
position = ftell (my_stream);
input_char = getc (my_stream);
printf ("Character at position %d = ’%c’.\n\n", position, input_char);

printf ("Seeking position 25...\n");

fseek (my_stream, 25, SEEK_SET);
position = ftell (my_stream);
input_char = getc (my_stream);
printf ("Character at position %d = ’%c’.\n\n", position, input_char);

printf ("Attempting to read again...\n");

input_char = getc (my_stream);
eof_status = feof (my_stream);
printf ("feof returns %d.\n\n", eof_status);

error_status = ferror (my_stream);

printf ("ferror returns %d.\n", error_status);
printf ("Attempting to write to this read-only stream...\n");
putc (’!’, my_stream);
error_status = ferror (my_stream);
printf ("ferror returns %d.\n\n", error_status);
putc and fputc 133

printf ("Rewinding...\n");
rewind (my_stream);
position = ftell (my_stream);
input_char = getc (my_stream);
printf ("Character at position %d = ’%c’.\n", position, input_char);

close_error = fclose (my_stream);

/* Handle fclose errors */

if (close_error != 0)
{
printf ("File could not be closed.\n");
}
else
{
printf ("File closed.\n");
}

return 0;
}

There is another function in the GNU C Library called fgetc. It is identical to getc in
most respects, except that getc is usually implemented as a macro function and is highly
optimised, so is preferable in most situations. (In situations where you are reading from
standard input, getc is about as fast as fgetc, since humans type slowly compared to how
fast computers can read their input, but when you are reading from a stream that is not
interactively produced by a human, fgetc is probably better.)

16.3.4 putc and fputc

If you want to write a single character to a stream other than stdout, you can use the putc
function. This function is very similar to putchar, but accepts an argument that specifies
the stream to which to write. It takes a single integer parameter containing a character (the
argument can be a single-quoted text character, as in the example below), and sends the
character to the specified stream. If a write error occurs, putc returns EOF; otherwise, it
returns the integer it was passed. This can simply be disregarded, as in the example below.

The following code example creates a text file called ‘snazzyjazz.txt’. It then writes
an ‘X’, a space, and then a line of ten exclamation marks (‘!!!!!!!!!!’) to the file, and
a newline character to it using the putc function. Notice the use of the for loop; by this
means, putchar can be used not just for one character, but multiple times. , then writes
ten exclamation mark characters (‘!!!!!!!!!!’)

#include <stdio.h>

int main()
{
int i;
FILE *my_stream;
char my_filename[] = "snazzyjazz.txt";

my_stream = fopen (my_filename, "w");

134 Chapter 16: Input and output

putc (’X’, my_stream);

putc (’ ’, my_stream);
for (i=1; i<=10; i++)
{
putc (’!’, my_stream);
}
putc (’\n’, my_stream);

/* Close stream; skip error-checking for brevity of example */

fclose (my_stream);

return 0;
}

There is another function in the GNU C Library called fputc. It is identical to putc in
most respects, except that putc is usually implemented as a macro function and is highly
optimised, so is preferable in most situations.

16.3.5 ungetc()
Every time a character is read from a stream by a function like getc, the file position
indicator advances by 1. It is possible to reverse the motion of the file position indicator
with the function ungetc, which steps the file position indicator back by one byte within
the file and reverses the effect of the last character read operation. (This is called unreading
the character or pushing it back onto the stream.)
The intended purpose is to leave the indicator in the correct file position when other
functions have moved too far ahead in the stream. Programs can therefore peek ahead, or
get a glimpse of the input they will read next, then reset the file position with ungetc.
On GNU systems, you cannot call ungetc twice in a row without reading at least one
character in between; in other words, GNU only supports one character of pushback.
Pushing back characters does not change the file being accessed at all; ungetc only
affects the stream buffer, not the file. If fseek, rewind, or some other file positioning
function is called, any character due to be pushed back by ungetc is discarded.
Unreading a character on a stream that is at end-of-file resets the end-of-file indicator
for the stream, because there is once again a character available to be read. However, if the
character pushed back onto the stream is EOF, ungetc does nothing and just returns EOF.
Here is a code example that reads all the whitespace at the beginning of a file with
getc, then backs up one byte to the first non-whitespace character, and reads all following
characters up to a newline character with the getline function. (See Section 16.2.6.1
[getline], page 120, for more information on that function.)
#include <stdio.h>

int main()
{
int in_char;
FILE *my_stream;
char *my_string = NULL;
size_t nchars = 0;
Programming with pipes 135

my_stream = fopen ("snazzyjazz.txt", "w");

fprintf (my_stream, " Here’s some non-whitespace.\n");

/* Close stream; skip error-checking for brevity of example */

fclose (my_stream);

my_stream = fopen ("snazzyjazz.txt", "r");

/* Skip all whitespace in stream */

do
in_char = getc (my_stream);
while (isspace (in_char));

/* Back up to first non-whitespace character */

ungetc (in_char, my_stream);

getline (&my_string, &nchars, my_stream);

/* Close stream; skip error-checking for brevity of example */

fclose (my_stream);

printf ("String read:\n");

printf ("%s", my_string);

return 0;
}
The code example will skip all initial whitespace in the file ‘snazzyjazz.txt’, and display
the following text on standard output:
String read:
Here’s some non-whitespace.

16.4 Programming with pipes

There may be times when you will wish to manipulate other programs on a GNU system
from within your C program. One good way to do so is the facility called a pipe. Using
pipes, you can read from or write to any program on a GNU system that writes to standard
output and reads from standard input. (In the ancestors of modern GNU systems, pipes
were frequently files on disk; now they are usually streams or something similar. They are
called “pipes” because people usually visualise data going in at one end and coming out at
the other.)
For example, you might wish to send output from your program to a printer. As men-
tioned in the introduction to this chapter, each printer on your system is assigned a device
name such as ‘/dev/lp0’. Pipes provide a better way to send output to the printer than
writing directly to the device, however.
Pipes are useful for many things, not just sending output to the printer. Suppose you
wish to list all programs and processes running on your computer that contain the string
‘init’ in their names. To do so at the GNU/Linux command line, you would type something
like the following command:
ps -A | grep init
This command line takes the output of the ps -A command, which lists all running
processes, and pipes it with the pipe symbol (‘|’) to the grep init command, which returns
136 Chapter 16: Input and output

all lines that were passed to it that contain the string ‘init’. The output of this whole
process will probably look something like this on your system:
1 ? 00:00:11 init
4884 tty6 00:00:00 xinit
The pipe symbol ‘|’ is very handy for command-line pipes and pipes within shell scripts,
but it is also possible to set up and use pipes within C programs. The two main C functions
to remember in this regard are popen and pclose.
The popen function accepts as its first argument a string containing a shell command,
such as lpr. Its second argument is a string containing either the mode argument ‘r’ or
‘w’. If you specify ‘r’, the pipe will be open for reading; if you specify ‘w’, it will be open
for writing. The return value is a stream open for reading or writing, as the case may be;
if there is an error, popen returns a null pointer.
The pclose function closes a pipe opened by popen. It accepts a single argument, the
stream to close. It waits for the stream to close, and returns the status code returned by
the program that was called by popen.
If you open the pipe for reading or writing, in between the popen and pclose calls, it
is possible to read from or write to the pipe in the same way that you might read from or
write to any other stream, with high-level input/output calls such as getdelim, fprintf
and so on.
The following program example shows how to pipe the output of the ps -A command to
the grep init command, exactly as in the GNU/Linux command line example above. The
output of this program should be almost exactly the same as sample output shown above.
#include <stdio.h>
#include <stdlib.h>

int
main ()
{
FILE *ps_pipe;
FILE *grep_pipe;

int bytes_read;
int nbytes = 100;
char *my_string;

/* Open our two pipes */

ps_pipe = popen ("ps -A", "r");
grep_pipe = popen ("grep init", "w");

/* Check that pipes are non-null, therefore open */

if ((!ps_pipe) || (!grep_pipe))
{
fprintf (stderr,
"One or both pipes failed.\n");
return EXIT_FAILURE;
}

/* Read from ps_pipe until two newlines */

my_string = (char *) malloc (nbytes + 1);
bytes_read = getdelim (&my_string, &nbytes, "\n\n", ps_pipe);
Usual file name errors 137

/* Close ps_pipe, checking for errors */

if (pclose (ps_pipe) != 0)
{
fprintf (stderr,
"Could not run ’ps’, or other error.\n");
}

/* Send output of ’ps -A’ to ’grep init’, with two newlines */

fprintf (grep_pipe, "%s\n\n", my_string);

/* Close grep_pipe, cehcking for errors */

if (pclose (grep_pipe) != 0)
{
fprintf (stderr,
"Could not run ’grep’, or other error.\n");
}

/* Exit! */
return 0;
}

16.5 Low-level file routines

High-level file routines such as those already described are usually convenient and easy to
use. However, low-level file routines such as the ones in this section have some advantages.
For example, they do not treat all file input/output as text streams, as the high-level routines
do; for that reason, working with binary files may be easier using low-level routines. For
another thing, low-level routines do not buffer their input and output, so you will never
need to remember to flush your streams with fflush or similar functions, as you sometimes
must with high-level routines.
Unfortunately, because low-level routines work at a lower level of abstraction, they can
be tricky, even dangerous to use — that is to say, if used incorrectly, they may corrupt your
data or cause your program to terminate unexpectedly; never fear, they will not explode
your monitor or cause your computer to become sapient and attempt world domination.
As mentioned, low-level file routines do not use text streams; instead, the connection
they open to your file is an integer called a file descriptor. You pass the file descriptor
that designates your file to most low-level file routines, just as you pass the stream that
designates your file to most high-level file routines. For example, while the low-level open
function takes a filename string to open a file, the matched close function takes the file
descriptor returned by open:
my_file_descriptor = open ("foo_file", O_RDONLY);
close_err = close (my_file_descriptor);

16.5.1 Usual file name errors

Most low-level file functions return some kind of error flag if they cannot perform the action
you request, for example, if they cannot parse the file name or find the file. However, to
discover which error or what kind of error has occurred, you must frequently refer to the
system variable errno. This is an integer specifying the most recent error that has occurred.
Macros for values of errno are listed below. They are all defined in the GNU C Library.
138 Chapter 16: Input and output

The word component below refers to part of a full file name. For example, in the file
name ‘/home/fred/snozzberry.txt’, ‘fred’ is a component that designates a subdirectory
of the directory ‘/home’, and ‘snozzberry.txt’ is the name of the file proper.
Most functions that accept file name arguments can detect the following error conditions.
These are known as the usual file name errors. The names of the errors, such as EACCES,
are compounded of ‘E’ for “error” and a term indicating the type of error, such as ‘ACCES’
for “access”.
EACCES The program is not permitted to search within one of the directories in the file
name.
ENAMETOOLONG
Either the full file name is too long, or some component is too long. GNU does
not limit the overall length of file names, but depending on which file system you
are using, the length of component names may be limited. (For example, you
may be running GNU/Linux but accessing a Macintosh HFS disk; the names
of Macintosh files cannot be longer than 31 characters.)
ENOENT Either some component of the file name does not exist, or some component is
a symbolic link whose target file does not exist.
ENOTDIR One of the file name components that is supposed to be a directory is not a
directory.
ELOOP Too many symbolic links had to be followed to find the file. (GNU has a limit
on how many symbolic links can be followed at once, as a basic way to detect
recursive (looping) links.)
You can display English text for each of these errors with the ‘m’ conversion specifier of
the printf function, as in the following short example.
errno = EACCES;
printf ("errno string (EACCES): %m\n");
This example prints the following string:
errno string (EACCES): Permission denied
See Section 16.2.2.2 [Formatted output conversion specifiers], page 115, for more information
on the ‘m’ conversion specifier.

16.5.2 Opening files at a low level

You can open a file, or create one if it does not already exist, with the open command,
which creates and returns a new file descriptor for the file name passed to it. If the file is
successfully opened, the file position indicator is initially set to zero (the beginning of the
file). (Note that the open function is actually called at an underlying level by fopen.)
The first parameter of open is a string containing the filename of the file you wish to
open. The second parameter is an integer argument created by the bitwise OR of the
following file status flags. (Bitwise OR is a mathematical operator that we have not yet
covered in this book. To perform bitwise OR on two variables a and b, you simply insert a
pipe character between them, thus: a | b. Bitwise OR is similar to the way the expression
“and/or” is used in English. See the code example below for the use of bitwise OR with file
status flags. See Chapter 18 [Advanced operators], page 167, for a detailed explanation of
bitwise OR and other bitwise operators.)
Opening files at a low level 139

The following flags are the more important ones for a beginning C programmer to know.
There are a number of file status flags which are relevant only to more advanced program-
mers; for more details, see section “File Status Flags” in The GNU C Library Reference
Manual.)
Note that these flags are defined in macros in the GNU C Library header file ‘fcntl.h’,
so remember to insert the line #include <fcntl.h> at the beginning of any source code
file that uses them.
O_RDONLY Open the file for read access.
O_WRONLY Open the file for write access.
O_RDWR Open the file for both read and write access. Same as O_RDONLY | O_WRONLY.
O_READ
Same as O_RDWR. GNU systems only.
O_WRITE Same as O_WRONLY. GNU systems only.
O_EXEC Open the file for executing. GNU systems only.
O_CREAT The file will be created if it doesn’t already exist.
O_EXCL If O_CREAT is set as well, then open fails if the specified file exists already. Set
this flag if you want to ensure you will not clobber an existing file.
O_TRUNC Truncate the file to a length of zero bytes. This option is not useful for direc-
tories or other such special files. You must have write permission for the file,
but you do not need to open it for write access to truncate it (under GNU).
O_APPEND Open the file for appending. All write operations then write the data at the
end of the file. This is the only way to ensure that the data you write will always
go to the end of the file, even if there are other write operations happening at
the same time.
The open function normally returns a non-negative integer file descriptor connected to
the specified file. If there is an error, open will return -1 instead. In that case, you can
check the errno variable to see which error occurred. In addition to the usual file name
errors, open can set errno to the following values. (It can also specify a few other errors of
interest only to advanced C programmers. See section “Opening and Closing Files” in The
GNU C Library Reference Manual, for a full list of error values. See Section 16.5.1 [Usual
file name errors], page 137, for a list of the usual file name errors.).
EACCES The file exists but is cannot be does not have read or write access (as requested),
or the file does not exist but cannot be created because the directory does not
have write access.
EEXIST Both O_CREAT and O_EXCL are set, and the named file already exists. To open
it would clobber it, so it will not be opened.
EISDIR Write access to the file was requested, but the file is actually a directory.
EMFILE Your program has too many files open.
ENOENT The file named does not exist, and O_CREAT was not specified, so the file will
not be created.
140 Chapter 16: Input and output

ENOSPC The file cannot be created, because the disk is out of space.
EROFS The file is on a read-only file system, but either one of O_WRONLY, O_RDWR, or
O_TRUNC was specified, or O_CREAT was set and the file does not exist.

See Section 16.5.3 [Closing files at a low level], page 140, for a code example using both
the low-level file functions open and close.

16.5.2.1 File creation

In older C code using low-level file routines, there was a function called creat that was
used for creating files. This function is still included in GNU for compatibility with older
C code, but is considered obsolete. In order to create a file, instead of writing
creat (filename )

it is now considered better coding to practice to write the following code:

open (filename, O_WRONLY | O_CREAT | O_TRUNC)

16.5.3 Closing files at a low level

To close a file descriptor, use the low-level file function close. The sole argument to close
is the file descriptor you wish to close.
The close function returns 0 if the call was successful, and -1 if there was an error.
In addition to the usual file name error codes, it can set the system variable errno to one
of the following values. It can also set errno to several other values, mostly of interest to
advanced C programmers. See section “Opening and Closing Files” in The GNU C Library
Reference Manual, for more information.

EBADF The file descriptor passed to close is not valid.

Remember, close a stream by using fclose instead. This allows the necessary system
bookkeeping to take place before the file is closed.
Here is a code example using both the low-level file functions open and close.
#include <stdio.h>
#include <fcntl.h>

int main()
{

char my_filename[] = "snazzyjazz17.txt";

int my_file_descriptor, close_err;

/*
Open my_filename for writing. Create it if it does not exist.
Do not clobber it if it does.
*/

my_file_descriptor = open (my_filename, O_WRONLY | O_CREAT | O_EXCL);

if (my_file_descriptor == -1)
{
printf ("Open failed.\n");
}
Writing files at a low level 141

close_err = close (my_file_descriptor);

if (close_err == -1)
{
printf ("Close failed.\n");
}

return 0;
}

Running the above code example for the first time should produce no errors, and should
create an empty text file called ‘snazzyjazz17.txt’. Running it a second time should
display the following errors on your monitor, since the file ‘snazzyjazz17.txt’ already
exists, and should not be clobbered according to the flags passed to open.
Open failed.
Close failed.

16.5.4 Reading files at a low level

You can read a block of information from a file with the read function. The data read is
loaded directly into a buffer in memory. The data can be binary as well as a text, but if the
latter, no terminating newline is added. The bytes read start at the current file position;
after reading them, read advances the file position to immediately after the bytes read.
The read function takes three parameters. The first one is the file descriptor from which
data is to be read. The second is the buffer in memory where the data read will be stored.
The buffer is of type void *, and can be an array or a chunk of space reserved with malloc.
The final parameter is of type size_t, and specifies the number of bytes to read.
The return value of this function is of type ssize_t, and represents the number of bytes
actually read. This might be less than the number of bytes requested if there are not
enough bytes left in the file or immediately available. Reading less than the number of
bytes requested does not generate an error.
If the number of bytes requested is not zero, a return value of zero indicates the end of
the file. This is also not an error. If you keep calling read at the end of the file, it will
simply keep returning zero. If read returns at least one character, you cannot tell whether
the end of the file was reached from that information, but read will return zero on the next
read operation if it was.
If there was an error, read returns -1. You can then check the system variable errno
for one of the following error conditions, as well as the usual file name errors. (See Sec-
tion 16.5.1 [Usual file name errors], page 137.) The read function can also return some
other error conditions in errno that are mostly of interest to advanced C programmers.
(See section “Input and Output Primitives” in The GNU C Library Reference Manual, for
more information.)

EBADF The file descriptor passed to read is not valid, or is not open for reading.
EIO There was a hardware error. (This error code also applies to more abstruse
conditions detailed in the GNU C Library manual.)

See Section 16.5.5 [Writing files at a low level], page 142, for a code example that uses
the read function.
142 Chapter 16: Input and output

16.5.5 Writing files at a low level

You can write a block of information to a file with the write function, which is called by
all high-level file writing routines, such as fwrite. It takes three parameters. The first is
the file descriptor of the file you wish to write to. The second is a buffer, of type void *,
that contains the data you wish to write. (It can be an array of bytes, but need not be a
text string. Null characters in the data are treated in the same way as other characters.)
The third parameter is of type size_t, and specifies the number of bytes that are to be
written.
The return value of this function is of type ssize_t, and indicates the number of bytes
actually written. This may be the same as the third parameter (the number of bytes to be
written), but may be less; you should always call write in a loop, and iterate the loop until
all data has been written. If there is an error, write returns -1. The write function will
return the following error codes in the system variable errno, as well as the usual file name
errors. (See Section 16.5.1 [Usual file name errors], page 137.)
EBADF The file descriptor specified is invalid, or is not open for writing.
EFBIG If the data were written, the file written to would become too large.
EIO There has been a hardware error.
EINTR The write operation was temporarily interrupted.
ENOSPC The device containing the file is full.
In addition to the error codes above, write can return some error codes that are mainly
of interest to advanced C programmers. If write fails, you should check errno to see if the
error was EINTR; if it was, you should repeat the write call each time.
Even though low-level file routines do not use buffering, and once you call write, your
data can be read from the file immediately, it may take up to a minute before your data
is physically written to disk. You can call the fsync routine (see below) to ensure that
all data is written to the file; this usage is roughly analogous to the high-level file routine
fflush.
The fsync routine takes a single parameter, the file descriptor to synchronise. It does
not return a value until all data has been written. If no error occurred, it returns a 0;
otherwise, it returns -1 and sets the system variable errno to one of the following values:
EBADF The file descriptor specified is invalid.
EINVAL No synchronization is possible because the system does not implement it.
Here is a code example that demonstrates the use of the write, read, and fsync func-
tions. (See Section 16.5.4 [Reading files at a low level], page 141, for more information on
read.)
#include <stdio.h>
#include <fcntl.h>

int main()
{
char my_write_str[] = "1234567890";
char my_read_str[100];
char my_filename[] = "snazzyjazz.txt";
int my_file_descriptor, close_err;
Finding file positions at a low level 143

/* Open the file. Clobber it if it exists. */

my_file_descriptor = open (my_filename, O_RDWR | O_CREAT | O_TRUNC);

/* Write 10 bytes of data and make sure it’s written */

write (my_file_descriptor, (void *) my_write_str, 10);
fsync (my_file_descriptor);

/* Seek the beginning of the file */

lseek (my_file_descriptor, 0, SEEK_SET);

/* Read 10 bytes of data */

read (my_file_descriptor, (void *) my_read_str, 10);

/* Terminate the data we’ve read with a null character */

my_read_str[10] = ’\0’;

printf ("String read = %s.\n", my_read_str);

close (my_file_descriptor);

return 0;
}

16.5.6 Finding file positions at a low level

If you want to find a particular file position within a file, using a low-level file routine, you
can call the lseek function. This is very similar to the high-level file routine fseek, except
that it accepts a file descriptor rather than a stream as an argument.
The lseek function specifies the file position for the next read or write operation. (See
Section 16.1.4 [File position], page 109, for more information on file positions.)
The lseek function takes three parameters. The first parameter is the file descriptor.
The second is of type off_t and specifies the number of bytes to move the file position
indicator. The third argument and the third parameter is a constant that specifies whether
the offset is relative to the beginning of the file (SEEK_SET), to the current file position
(SEEK_CUR), or to the end of the file (SEEK_END). If SEEK_CUR or SEEK_END is used, the
offset specified can be positive or negative. If you specify SEEK_END, set the position past
the current end of the file, and actually write data, you will extend the file with zeroes up
to the position you specify. However, the blocks of zeroes are not actually written to disk,
so the file takes up less space on disk than it seems to; this is called a sparse file.
The return value of lseek is of type off_t and normally contains the resulting file
position, as measured in bytes from the beginning of the file. If you wish to read the
current file position, therefore, you can specify an offset of 0 and a third parameter of
SEEK_CUR, as follows:
file_position = lseek (file_descriptor, 0, SEEK_CUR);
If there was an error, lseek returns a -1 and sets the system variable errno to one of
the following values:
EBADF The file descriptor specified is invalid.
EINVAL Either the third parameter of lseek is invalid, or the file offset is invalid.
ESPIPE The file descriptor corresponds to an object that cannot be positioned, such as
a terminal device.
144 Chapter 16: Input and output

The lseek function is called by many high-level file position functions, including fseek,
rewind, and ftell.

16.5.7 Deleting files at a low level

If you want to delete a file, you can use the low-level file routine unlink, as declared in the
file ‘unistd.h’. Simply pass this routine the name of the file you wish to delete. If this is
the only name the file has (that is, if no one has created a hard link to the file with the link
function, the GNU command ln, or something similar), then the file itself will be deleted;
otherwise, only that name will be deleted. (See the section "Hard Links" in the GNU C
Library manual for more information on hard links.) If the file is open when unlink is
called, unlink will wait for the file to be closed before it deletes it.
The unlink function returns 0 if the file or file name was successfully deleted. If there
was an error, unlink returns -1. In addition to the usual file name errors, unlink can set
errno to the following values. (See Section 16.5.1 [Usual file name errors], page 137, for a
list of the usual file name errors.)
EACCES Your program does not have permission to delete the file from the directory
that contains it.
EBUSY The file is currently being used by the system and cannot be deleted.
ENOENT The file name to be deleted does not exist.
EPERM Your program tried to delete a directory with unlink; this is not permitted
under GNU. (See remove below.)
EROFS The file name is on a read-only file system and cannot be deleted.
If you wish to delete a directory rather than an ordinary file, use the rmdir function.
Simply pass it the name of an empty directory you wish to delete. It acts like unlink in
most respects, except that it can return an extra error code in the system variable errno:
ENOTEMPTY
The directory was not empty, so cannot be deleted. This code is synonymous
with EEXIST, but GNU always returns ENOTEMPTY.

16.5.8 Renaming files at a low level

If you want to rename a file, you can use the rename function, which takes two parameters.
The first parameter is a string containing the old name of the file, and the second is a string
containing the new name. (As with unlink, this function only operates on one of the names
of a file, if the file has hard links. See Section 16.5.7 [Deleting files at a low level], page 144,
for caveats and information on hard links.)
Both the new name and the old name must be on the same file system. Any file in the
same directory that has the same name as the new file name will be deleted in the process
of renaming the file.
If rename fails, it will return -1. In addition to the usual file name errors, unlink can
set errno to the following values. (See Section 16.5.1 [Usual file name errors], page 137, for
a list of the usual file name errors.)
EACCES Either one of the directories in question (either the one containing the old name
or the one containing the new name) refuses write permission, or the new name
Questions 145

and the old name are directories, and write permission is refused for at least
one of them.
EBUSY One of the directories used by the old name or the new name is being used by
the system and cannot be changed.
ENOTEMPTY
The directory was not empty, so cannot be deleted. This code is synonymous
with EEXIST, but GNU always returns ENOTEMPTY.
EINVAL The old name is a directory that contains the new name.
EISDIR The new name is a directory, but the old name is not.
EMLINK The parent directory of the new name would contain too many entries if the
new name were created.
ENOENT The old name does not exist.
ENOSPC The directory that would contain the new name has no room for another entry,
and cannot be expanded.
EROFS The rename operation would involve writing on a read-only file system.
EXDEV The new name and the old name are on different file systems.

16.6 Questions
1. What are the following?
1. File name
2. File descriptor
3. Stream
2. What is a pseudo-device name?
3. Where does ‘stdin’ usually get its input?
4. Where does ‘stdout’ usually send its output?
5. Write a program that simply prints out the following string to the screen: ‘6.23e+00’.
6. Investigate what happens when you type the wrong conversion specifier in a program.
e.g. try printing an integer with ‘%f’ or a floating point number with ‘%c’. This is
bound to go wrong – but how will it go wrong?
7. What is wrong with the following statements?
1. printf (x);
2. printf ("%d");
3. printf ();
4. printf ("Number = %d");
Hint: if you don’t know, try them in a program!
8. What is a whitespace character?
9. Write a program that aceepts two integers from the user, multiplies them together, and
prints the answer on your printer. Try to make the input as safe as possible.
146 Chapter 16: Input and output

10. Write a program that simply echoes all the input to the output.
11. Write a program that strips all space characters out of the input and replaces each
string of them with a single newline character.
12. The scanf function always takes pointer arguments. True or false?
13. What is the basic difference between high-level and low-level file routines?
14. Write a statement that opens a high level file for reading.
15. Write a statement that opens a low level file for writing.
16. Write a program that checks for illegal characters in text files. The only valid characters
are ASCII codes 10, 13, and 32..126.
17. What statement performs formatted writing to text files?
18. Poke around in the header files on your system so you can see what is defined where.
argc and argv 147

17 Putting a program together

This chapter explains, step by step, how to create a “real” program that meets GNU
standards for a command-line interface. It also discusses how to create a program whose
source is split into multiple files, and how to compile it, with or without the GNU utility
make. Finally, it discusses how to create a code library, in case you write some useful
functions that you want to share with other programmers.
Putting it all together.

17.1 argc and argv

So far, all the programs we have written can be run with a single command. For example,
if we compile an executable called ‘myprog’, we can run it from within the same directory
with the following command at the GNU/Linux command line:
./myprog
However, what if you want to pass information from the command line to the program
you are running? Consider a more complex program like GCC. To compile the hypothetical
‘myprog’ executable, we type something like the following at the command line:
gcc -o myprog myprog.c
The character strings ‘-o’, ‘myprog’, and ‘myprog.c’ are all arguments to the gcc com-
mand. (Technically ‘gcc’ is an argument as well, as we shall see.)
Command-line arguments are very useful. After all, C functions wouldn’t be very useful
if you couldn’t ever pass arguments to them — adding the ability to pass arguments to
programs makes them that much more useful. In fact, all the arguments you pass on the
command line end up as arguments to the main function in your program.
Up until now, the skeletons we have used for our C programs have looked something like
this:
#include <stdio.h>

int main()
{

return 0;
}
From now on, our examples may look a bit more like this:
#include <stdio.h>

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

{

return 0;
}
As you can see, main now has arguments. The name of the variable argc stands for
“argument count”; argc contains the number of arguments passed to the program. The
name of the variable argv stands for “argument vector”. A vector is a one-dimensional
array, and argv is a one-dimensional array of strings. Each string is one of the arguments
that was passed to the program.
For example, the command line
148 Chapter 17: Putting a program together

gcc -o myprog myprog.c

would result in the following values internal to GCC:

argc 4
argv[0] ‘gcc’
argv[1] ‘-o’
argv[2] ‘myprog’
argv[3] ‘myprog.c’
As you can see, the first argument (argv[0]) is the name by which the program was
called, in this case ‘gcc’. Thus, there will always be at least one argument to a program,
and argc will always be at least 1.
The following program accepts any number of command-line arguments and prints them
out:
#include <stdio.h>

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

{
int count;

printf ("This program was called with \"%s\".\n",argv[0]);

if (argc > 1)
{
for (count = 1; count < argc; count++)
{
printf("argv[%d] = %s\n", count, argv[count]);
}
}
else
{
printf("The command had no other arguments.\n");
}

return 0;
}
If you name your executable ‘fubar’, and call it with the command ‘./fubar a b c’, it
will print out the following text:
This program was called with "./fubar".
argv[1] = a
argv[2] = b
argv[3] = c

17.2 Processing command-line options

It is easy, though tedious to pull options directly out of the argv vector with your own
routines. It is slightly less tedious to use the standard C option-processing function getopt,
or the enhanced GNU version of the same function, getopt_long, which permits GNU-style
long options (for example, ‘--quiet’ as opposed to ‘-q’).
The best option of all is to use the argp interface for processing options. Professionally
written programs provide the user with standard and useful options. The argp function
argp description 149

provides for these. For the modest price of setting up your command line arguments in a
structured way, and with surprisingly few lines of code, you can obtain all the perks of a
“real” GNU program, such as “automagically”-generated output to the ‘--help’, ‘--usage’,
and ‘--version’ options, as defined by the GNU coding standards. Using argp results in
a more consistent look-and-feel for programs that use it, and makes it less likely that the
built-in documentation for a program will be wrong or out of date.
POSIX, the Portable Operating System Interface standard, recommends the following
conventions for command-line arguments. The argp interface makes implementing them
easy.
• Command-line arguments are options if they begin with a hyphen (‘-’).
• Multiple options may follow a hyphen in a cluster if they do not take arguments. Thus,
‘-abc’ and ‘-a -b -c’ are the same.
• Option names are single alphanumeric characters.
• Options may require an argument. For example, the ‘-o’ option of the ld command
requires an output file name.
• The whitespace separating an option and its argument is optional. Thus, ‘-o foo’ and
‘-ofoo’ are the same.
• Options usually precede non-option arguments. (In fact, argp is more flexible than this;
if you want to suppress this flexibility, define the _POSIX_OPTION_ORDER environment
variable.)
• The argument ‘--’ terminates all options; all following command-line arguments are
considered non-option arguments, even if they begin with a hyphen.
• A single hyphen as an argument is considered a non-option argument; by convention,
it is used to specify input from standard input or output to standard output.
• Options may appear in any order, even multiple times. The meaning of this is left to
the application.
In addition, GNU adds long options, like the ‘--help’, ‘--usage’, and ‘--version’
options mentioned above. A long option starts with ‘--’, which is then followed by a string
of alphanumeric characters and hyphens. Option names are usually one to three words
long, with hyphens to separate words. Users can abbreviate the option names as long as
the abbreviations are unique. A long option (such as ‘--verbose’) often has a short-option
synonym (such as ‘-v’).
Long options can accept optional (that is, non-necessary) arguments. You can specify
an argument for a long option as follows:
‘--’option-name ‘=’value
You may not type whitespace between the option name and the equals sign, or between the
equals sign and the option value.

17.2.1 argp description

This section will describe how to write a simple program that implements most of the
standards mentioned above. It assumes some knowledge of advanced C data structures
that we have not yet covered in this book; if you are confused, you might want to consult
the chapter that discusses this material. (See Chapter 19 [More data types], page 177.)
Note that we are only discussing the basics of argp in this chapter; to read more about this
150 Chapter 17: Putting a program together

complicated and flexible facility of the GNU C Library, consult section “Parsing Program
Options with Argp” in The GNU C Library Reference Manual. Nevertheless, what you
learn in this chapter may be all you need to develop a program that is compliant with GNU
coding standards, with respect to command-line options.
The main interface to argp is the argp_parse function. Usually, the only argument-
parsing code you will need in main is a call to this function. The first parameter it takes
is of type const struct argp *argp, and specifies an ARGP structure (see below). (A value
of zero is the same as a structure containing all zeros.) The second parameter is simply
argc, the third simply argv. The fourth parameter is a set of flags that modify the parsing
behaviour; setting this to zero usually doesn’t hurt unless you’re doing something fancy, and
the same goes for the fifth parameter. The sixth parameter can be useful; in the example
below, we use it to pass information from main to our function parse_opt, which does
most of the work of initalizing internal variables (fields in the arguments structure) based
on command-line options and arguments.
The argp_parse returns a value of type error_t: usually either 0 for success, ENOMEM if
a memory allocation error occurred, or EINVAL if an unknown option or argument was met
with.
For this example, we are using only the first four fields in ARGP, which are usually all
that is needed. The rest of the fields will default to zero. The four fields are, in order:
1. OPTIONS: A pointer to a vector the elements of which are of type struct argp_option,
which contains four fields. The vector elements specify which options this parser un-
derstands. If you assign your option structure by initializing the array as we do in
this section’s main example, unspecified fields will default to zero, and need not be
specified. The whole vector may contain zero if there are no options at all. It should
in any case be terminated by an entry with a zero in all fields (as we do by specifying
the last item in the options vector to be {0} in the main example below.
The four main argp_option structure fields are as follows. (We will ignore the fifth
one, which is relatively unimportant and will simply default to zero if you do not specify
it.)
1. NAME: The name of this option’s long option (may be zero). To specify multiple
names for an option, follow it with additional entries, with the OPTION_ALIAS flag
set.
2. KEY: The integer key to pass to the PARSER function when parsing the current
option; this is the same as the name of the current option’s short option, if it is a
printable ASCII character.
3. ARG: The name of this option’s argument, if any.
4. FLAGS: Flags describing this option. You can specify multiple flags with logical
OR (for example, OPTION_ARG_OPTIONAL | OPTION_ALIAS).
Some of the available options are:
• OPTION_ARG_OPTIONAL: The argument to the current option is optional.
• OPTION_ALIAS: The current option is an alias for the previous option.
• OPTION_HIDDEN: Don’t show the current option in --help output.
5. DOC: A documentation string for the current option; will be shown in --help
output.
argp example 151

2. PARSER: A pointer to a function to be called by argp for each option parsed. It should
return one of the following values:
• 0: Success.
• ARGP_ERR_UNKNOWN: The given key was not recognized.
• An errno value indicating some other error. (See Section 16.5.1 [Usual file name
errors], page 137.)

The parser function takes the following arguments:

1. KEY: An integer specifying which argument this is, taken from the KEY field in each
argp_option structure, or else a key with a special meaning, such as one of the
following:
• ARGP_KEY_ARG: The current command-line argument is not an option.
• ARGP_KEY_END: All command-line arguments have been parsed.
2. ARG: The string value of the current command-line argument, or NULL if it has
none.
3. STATE: A pointer to an argp_state structure, containing information about the
parsing state, such as the following fields:
1. input: The same as the last parameter to argp_parse. We use this in the
main code example below to pass information between the main and parse_
opt functions.
2. arg_num: The number of the current non-option argument being parsed.
3. ARGS_DOC: If non-zero, a string describing how the non-option arguments should look.
It is only used to print the ‘Usage:’ message. If it contains newlines, the strings
separated by them are considered alternative usage patterns, and printed on separate
lines (subsequent lines are preceded by ‘or:’ rather than ‘Usage:’.
4. DOC: If non-zero, a descriptive string about this program. It will normally be printed
before the options in a help message, but if you include a vertical tab character (‘\v’),
the part after the vertical tab will be printed following the options in the output to
the --help option. Conventionally, the part before the options is just a short string
that says what the program does, while the part afterwards is longer and describes the
program in more detail.

There are also some utility functions associated with argp, such as argp_usage, which
prints out the standard usage message. We use this function in the parse_opt function in
the following example. See section “Functions For Use in Argp Parsers” in The GNU C
Library Reference Manual, for more of these utility functions.

17.2.2 argp example

Here is a code example that uses argp to parse command-line options. Remember, to
compile this example, copy it to a file called something like ‘argex.c’, then compile it
with the command gcc -o argex argex.c and run the resulting binary with the command
./argex.
152 Chapter 17: Putting a program together

#include <stdio.h>
#include <argp.h>

const char *argp_program_version =

"argex 1.0";

const char *argp_program_bug_address =

"<[email protected]>";

/* This structure is used by main to communicate with parse_opt. */

struct arguments
{
char *args[2]; /* ARG1 and ARG2 */
int verbose; /* The -v flag */
char *outfile; /* Argument for -o */
char *string1, *string2; /* Arguments for -a and -b */
};

/*
OPTIONS. Field 1 in ARGP.
Order of fields: {NAME, KEY, ARG, FLAGS, DOC}.
*/
static struct argp_option options[] =
{
{"verbose", ’v’, 0, 0, "Produce verbose output"},
{"alpha", ’a’, "STRING1", 0,
"Do something with STRING1 related to the letter A"},
{"bravo", ’b’, "STRING2", 0,
"Do something with STRING2 related to the letter B"},
{"output", ’o’, "OUTFILE", 0,
"Output to OUTFILE instead of to standard output"},
{0}
};

/*
PARSER. Field 2 in ARGP.
Order of parameters: KEY, ARG, STATE.
*/
static error_t
parse_opt (int key, char *arg, struct argp_state *state)
{
struct arguments *arguments = state->input;

switch (key)
{
case ’v’:
arguments->verbose = 1;
break;
case ’a’:
arguments->string1 = arg;
break;
case ’b’:
arguments->string2 = arg;
break;
case ’o’:
arguments->outfile = arg;
break;
argp example 153

case ARGP_KEY_ARG:
if (state->arg_num >= 2)
{
argp_usage(state);
}
arguments->args[state->arg_num] = arg;
break;
case ARGP_KEY_END:
if (state->arg_num < 2)
{
argp_usage (state);
}
break;
default:
return ARGP_ERR_UNKNOWN;
}
return 0;
}

/*
ARGS_DOC. Field 3 in ARGP.
A description of the non-option command-line arguments
that we accept.
*/
static char args_doc[] = "ARG1 ARG2";

/*
DOC. Field 4 in ARGP.
Program documentation.
*/
static char doc[] =
"argex -- A program to demonstrate how to code command-line options
and arguments.\vFrom the GNU C Tutorial.";

/*
The ARGP structure itself.
*/
static struct argp argp = {options, parse_opt, args_doc, doc};

/*
The main function.
Notice how now the only function call needed to process
all command-line options and arguments nicely
is argp_parse.
*/
int main (int argc, char **argv)
{
struct arguments arguments;
FILE *outstream;
154 Chapter 17: Putting a program together

char waters[] =
"a place to stay
enough to eat
somewhere old heroes shuffle safely down the street
where you can speak out loud
about your doubts and fears
and what’s more no-one ever disappears
you never hear their standard issue kicking in your door
you can relax on both sides of the tracks
and maniacs don’t blow holes in bandsmen by remote control
and everyone has recourse to the law
and no-one kills the children anymore
and no-one kills the children anymore
--\"the gunners dream\", Roger Waters, 1983\n";

/* Set argument defaults */

arguments.outfile = NULL;
arguments.string1 = "";
arguments.string2 = "";
arguments.verbose = 0;

/* Where the magic happens */

argp_parse (&argp, argc, argv, 0, 0, &arguments);

/* Where do we send output? */

if (arguments.outfile)
outstream = fopen (arguments.outfile, "w");
else
outstream = stdout;

/* Print argument values */

fprintf (outstream, "alpha = %s\nbravo = %s\n\n",
arguments.string1, arguments.string2);
fprintf (outstream, "ARG1 = %s\nARG2 = %s\n\n",
arguments.args[0],
arguments.args[1]);

/* If in verbose mode, print song stanza */

if (arguments.verbose)
fprintf (outstream, waters);

return 0;
}

Compile the code, then experiment! For example, here is the program output if you
simply type argex:

Usage: argex [OPTION...] ARG1 ARG2

Try ‘argex --help’ or ‘argex --usage’ for more information.

Here is the output from argex --usage:

Usage: argex [-v?V] [-a STRING1] [-b STRING2] [-o OUTFILE] [--alpha=STRING1]
[--bravo=STRING2] [--output=OUTFILE] [--verbose] [--help] [--usage]
[--version] ARG1 ARG2

Here is the output from argex --help:

Environment variables 155

Usage: argex [OPTION...] ARG1 ARG2

argex -- A program to demonstrate how to code command-line options
and arguments.

-a, --alpha=STRING1 Do something with STRING1 related to the letter A

-b, --bravo=STRING2 Do something with STRING2 related to the letter B
-o, --output=OUTFILE Output to OUTFILE instead of to standard output
-v, --verbose Produce verbose output
-?, --help Give this help list
--usage Give a short usage message
-V, --version Print program version

Mandatory or optional arguments to long options are also mandatory or optional

for any corresponding short options.

From the GNU C Tutorial.

Report bugs to <[email protected]>.

Here is the output from argex Foo Bar:
alpha =
bravo =

ARG1 = Foo
ARG2 = Bar

And finally, here is the output from argex --verbose -a 123 --bravo=456 Foo Bar:
alpha = 123
bravo = 456

ARG1 = Foo
ARG2 = Bar

a place to stay
enough to eat
somewhere old heroes shuffle safely down the street
where you can speak out loud
about your doubts and fears
and what’s more no-one ever disappears
you never hear their standard issue kicking in your door
you can relax on both sides of the tracks
and maniacs don’t blow holes in bandsmen by remote control
and everyone has recourse to the law
and no-one kills the children anymore
and no-one kills the children anymore
--"the gunners dream", Roger Waters, 1983
You can of course also send the output to a text file with the ‘-o’ or ‘--output’ option.

17.3 Environment variables

Sometimes it is useful to communicate with a program in a semi-permanent way, so that
you do not need to specify a command-line option every time you type the command to
execute the program. One way to do this is to generate a configuration file, in which you
can store data that will be used by the program every time it is run. This approach is
typically useful if you have a large amount of data that you want to pass to a program
every time it runs, or if you want the program itself to be able to change the data.
156 Chapter 17: Putting a program together

However, environment variables provide a more lightweight approach. Environment

variables, sometimes called shell variables, are usually set with the export command in the
shell. (This section assumes you are using the GNU Bash shell.) Standard environment
variables are used for information about your home directory, terminal type, and so on; you
can define additional variables for other purposes. The set of all environment variables that
have values is collectively known as the environment.
Names of environment variables are case-sensitive, and it is good form to use all upper-
case letters when defining a new variable; certainly this is the case for all system-defined
environment variables.
The value of an environment variable can be any string that does not contain a null
character (since the null character is used to terminate the string value).
Environment variables are stored in a special array that can be read by your main
function. Here is the skeleton for a main function that can read environment variables;
notice we have added a third parameter to main, called envp, which comes after argc and
argv.
#include <stdio.h>

/* To shorten example, not using argp */

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

return 0;
}
Notice that envp is an array of strings, just as argv is. It consists of a list of the
environment variables of your shell, in the following format:
NAME =value
Just as you can manually process command-line options from argv, so can you man-
ually process environment variables from envp. However, the simplest way to access the
value of an environment variable is with the getenv function, defined in the system header
‘stdlib.h’. It takes a single argument, a string containing the name of the variable whose
value you wish to discover. It returns that value, or a null pointer if the variable is not
defined.
#include <stdio.h>
#include <stdlib.h>

/* To shorten example, not using argp */

int main (int argc, char *argv[], char *envp[])
{
char *home, *host;

home = getenv("HOME");
host = getenv("HOSTNAME");

printf ("Your home directory is %s on %s.\n", home, host);

return 0;
}
When you run this code, it will print out a line like the following one.
Your home directory is /home/rwhe on linnaeus.
Writing a makefile 157

Note: Do not modify strings returned from getenv; they are pointers to data that
belongs to the system. If you want to process a value returned from getenv, copy it to
another string first with strcpy. (See Chapter 15 [Strings], page 97.) If you want to change
an environment variable from within your program (not usually advisable), use the putenv,
setenv, and unsetenv functions. See section “Environment Access” in The GNU C Library
Reference Manual, for more information on these functions.

17.4 Compiling multiple files

It is usually very simple to compile a program that has been divided across multiple source
files. Instead of typing
gcc -o executable sourcefile.c
you would type
gcc -o executable sourcefile_1.c sourcefile_2.c ... sourcefile_n.c
For example, if you were building a simple database program called ‘mydb’, the command
line might look something like this:
gcc -o mydb main.c keyboard_io.c db_access.c sorting.c
Of course, if (say) ‘db_access.c’ were lengthy, it might take a long time to compile your
program every time you executed this command, even if you only made a small change in
one of the other files. To avoid this, you might want to compile each of the source files into
its own object file, then link them together to make your program. If you did, each time
you made a small change in one file, you need only recompile that single file and then link
the object files together again, potentially a great savings in time and patience. Here is how
to generate a permanent object file for ‘db_access.c’.
gcc -c db_access.c
This would generate a permanent object code file called ‘db_access.o’, indicated by the
suffix ‘.o’. You would perform this step when needed for each of the source code files, then
link them together with the following command line:
gcc -o mydb main.o keyboard_io.o db_access.o sorting.o
You might even put the various commands into a shell file, so that you wouldn’t need
to type them repeatedly. For example, you could put the last command line into a shell file
called ‘build’, so that all you would need to do to build your executable from object code
files is type the following line.
./build
For programs on a very small scale, this approach works quite well. If your project grows
even slightly complex, however, you will have a hard time keeping track of which object
files are “fresh” and which need to be recreated because the corresponding source files have
been changed since their last compilation. That’s where the GNU utility make comes in.
(See Section 17.5 [Writing a makefile], page 157.)

17.5 Writing a makefile

The GNU make program automatically determines which pieces of a large program need to
be recompiled, and issues the commands to compile them. You need a file called a makefile
to tell make what to do. Most often, the makefile tells make how to compile and link a
program.
158 Chapter 17: Putting a program together

In this section, we will discuss a simple makefile that describes how to compile and link
a text editor which consists of eight C source files and three header files. The makefile can
also tell make how to run miscellaneous commands when explicitly asked (for example, to
remove certain files as a clean-up operation).
Although the examples in this section show C programs, you can use make with any
programming language whose compiler can be run with a shell command. Indeed, make
is not limited to programs. You can use it to describe any task where some files must be
updated automatically from others whenever the others change.
Your makefile describes the relationships among files in your program and provides
commands for updating each file. In a program, typically, the executable file is updated
from object files, which are in turn made by compiling source files.
Once a suitable makefile exists, each time you change some source files, this simple shell
command:
make
suffices to perform all necessary recompilations. The make program uses the makefile data-
base and the last-modification times of the files to decide which of the files need to be
updated. For each of those files, it issues the commands recorded in the database.
You can provide command line arguments to make to control which files should be
recompiled, or how.
When make recompiles the editor, each changed C source file must be recompiled. If a
header file has changed, each C source file that includes the header file must be recompiled to
be safe. Each compilation produces an object file corresponding to the source file. Finally,
if any source file has been recompiled, all the object files, whether newly made or saved
from previous compilations, must be linked together to produce the new executable editor.

17.5.1 What a Rule Looks Like

A simple makefile consists of “rules” with the following shape:
target ... : prerequisites ...
command
...
...
A target is usually the name of a file that is generated by a program; examples of targets
are executable or object files. A target can also be the name of an action to carry out, such
as ‘clean’.
A prerequisite is a file that is used as input to create the target. A target often depends
on several files.
A command is an action that make carries out. A rule may have more than one command,
each on its own line. Please note: you need to put a tab character at the beginning of every
command line! This is a common mistake that even experienced makefile writers can make.
Usually a command is defined by a rule with prerequisites and serves to create a target
file if any of the prerequisites change. However, the rule that specifies commands for the
target need not have prerequisites. For example, the rule containing the delete command
associated with the target ‘clean’ does not have prerequisites.
A simple makefile 159

A rule, then, explains how and when to remake certain files which are the targets of the
particular rule. make carries out the commands on the prerequisites to create or update the
target. A rule can also explain how and when to carry out an action.
A makefile may contain other text besides rules, but a simple makefile need only contain
rules. Rules may look somewhat more complicated than shown in this template, but all fit
the pattern more or less.

17.5.2 A simple makefile

Here is a straightforward makefile that describes the way an executable file called ‘edit’
depends on eight object files which, in turn, depend on eight C source files and three header
files.
In this example, all the C files include ‘defs.h’, but only files that define editing com-
mands include ‘command.h’, and only low-level files that change the editor buffer include
‘buffer.h’.
edit : main.o kbd.o command.o display.o \
insert.o search.o files.o utils.o
cc -o edit main.o kbd.o command.o display.o \
insert.o search.o files.o utils.o

main.o : main.c defs.h

cc -c main.c
kbd.o : kbd.c defs.h command.h
cc -c kbd.c
command.o : command.c defs.h command.h
cc -c command.c
display.o : display.c defs.h buffer.h
cc -c display.c
insert.o : insert.c defs.h buffer.h
cc -c insert.c
search.o : search.c defs.h buffer.h
cc -c search.c
files.o : files.c defs.h buffer.h command.h
cc -c files.c
utils.o : utils.c defs.h
cc -c utils.c
clean :
rm edit main.o kbd.o command.o display.o \
insert.o search.o files.o utils.o
We split each long line into two lines using a backslash; this is like using one long line, but
easier to read.
To use this makefile to create the executable file called ‘edit’, type:
make
To use this makefile to delete the executable file and all the object files from the directory,
type:
make clean
In the example makefile, the targets include the executable file ‘edit’, and the object
files ‘main.o’ and ‘kbd.o’. The prerequisites are files such as ‘main.c’ and ‘defs.h’. In
fact, each ‘.o’ file is both a target and a prerequisite. Commands include cc -c main.c
and cc -c kbd.c.
160 Chapter 17: Putting a program together

When a target is a file, it needs to be recompiled or relinked if any of its prerequisites

change. In addition, any prerequisites that are themselves automatically generated should
be updated first. In this example, ‘edit’ depends on each of the eight object files; the object
file ‘main.o’ depends on the source file ‘main.c’ and on the header file ‘defs.h’.
A shell command follows each line that contains a target and prerequisites. These shell
commands tell make how to update the target file. A tab character must come at the
beginning of every command line to distinguish command lines from other lines in the
makefile. (Bear in mind that make does not know anything about how the commands work.
It is up to you to supply commands that will update the target file properly.)
The target clean is not a file, but merely the name of an action. Since this action is
not carried out as part of the other targets, clean is not a prerequisite of any other rule.
Consequently, make never does anything with it unless you explicitly type make clean. Not
only is this rule not a prerequisite, it does not have any prerequisites itself, so the only
purpose of the rule is to run the specified commands. Targets like clean that do not refer
to files but are just actions are called phony targets.

17.5.3 make in action

By default, make starts with the first target whose name does not start with ‘.’. This is
called the default goal. (Goals are the targets that make tries to update.)
In the simple example of the previous section, the default goal is to update the executable
program ‘edit’; therefore, we put that rule first.
Thus, when you give the command:
make
make reads the makefile in the current directory and begins by processing the first rule. In
the example, this rule is for relinking ‘edit’; but before make can fully process this rule, it
must process the rules for the files that ‘edit’ depends on, which in this case are the object
files. Each of these files is processed according to its own rule. These rules say to update
each ‘.o’ file by compiling its source file. The recompilation must be done if the source file,
or any of the header files named as prerequisites, is more recent than the object file, or if
the object file does not exist.
The other rules are processed because their targets appear as prerequisites of the goal. If
some other rule is not depended on by the goal (or anything that the goal depends on, and
so forth), then that rule is not processed, unless you tell make to do so (with a command
such as make clean.
Before recompiling an object file, make considers updating its prerequisites (the source
file and header files). This makefile does not specify anything to be done for them—the
‘.c’ and ‘.h’ files are not the targets of any rules—so make does nothing for these files.
But make can update automatically generated C programs, such as those made by Bison or
Yacc, by defining the ‘.c’ and ‘.h’ files as targets and specifying how to create them with
Bison, Yacc, or whatever other program generated them.
After recompiling the appropriate object files, make decides whether to link ‘edit’. This
must be done if the file ‘edit’ does not exist, or if any of the object files are newer than it
is. If an object file was just recompiled, it is now newer than ‘edit’, so ‘edit’ is relinked.
Thus, if we change the file ‘insert.c’ and run make, then make will recompile that file,
update ‘insert.o’, and then link ‘edit’. If we change the file ‘command.h’ and run make,
Letting make deduce commands 161

make will recompile the object files ‘kbd.o’, ‘command.o’ and ‘files.o’, and then link the
file ‘edit’.

17.5.4 Variables simplify makefiles

In our example, we had to list all the object files twice in the rule for ‘edit’ (repeated here):
edit : main.o kbd.o command.o display.o \
insert.o search.o files.o utils.o
cc -o edit main.o kbd.o command.o display.o \
insert.o search.o files.o utils.o
Such duplication is error-prone; if a new object file is added to the system, we might add
it to one list and forget the other. We can eliminate the risk and simplify the makefile by
using a variable. Variables in make enable a text string to be defined once and substituted
in multiple places later. They are similar to C macros. (See Section 12.2 [Macros], page 68.)
It is standard practice for every makefile to have a variable named objects, OBJECTS,
objs, OBJS, obj, or OBJ that is a list of all object file names. We would define such a
variable objects with a line like this in the makefile:
objects = main.o kbd.o command.o display.o \
insert.o search.o files.o utils.o
Then, in every place we want to put a list of the object file names, we can substitute the
variable’s value by writing $(objects)
Here is how the complete simple makefile looks when you use a variable for the object
files:
objects = main.o kbd.o command.o display.o \
insert.o search.o files.o utils.o

edit : $(objects)
cc -o edit $(objects)
main.o : main.c defs.h
cc -c main.c
kbd.o : kbd.c defs.h command.h
cc -c kbd.c
command.o : command.c defs.h command.h
cc -c command.c
display.o : display.c defs.h buffer.h
cc -c display.c
insert.o : insert.c defs.h buffer.h
cc -c insert.c
search.o : search.c defs.h buffer.h
cc -c search.c
files.o : files.c defs.h buffer.h command.h
cc -c files.c
utils.o : utils.c defs.h
cc -c utils.c
clean :
rm edit $(objects)

17.5.5 Letting make deduce commands

It is not necessary to spell out the commands for compiling the individual C source files,
because make can figure them out: it has an implicit rule for updating a ‘.o’ file from
a correspondingly named ‘.c’ file using a gcc -c command. For example, it will use the
162 Chapter 17: Putting a program together

command gcc -c main.c -o main.o to compile ‘main.c’ into ‘main.o’. We can therefore
omit the commands from the rules for the object files.
When a ‘.c’ file is used automatically in this way, it is also automatically added to the
list of prerequisites. We can therefore omit the ‘.c’ files from the prerequisites, provided
we omit the commands.
Here is the entire example, with both of these changes, and the variable objects as
suggested above:
objects = main.o kbd.o command.o display.o \
insert.o search.o files.o utils.o

edit : $(objects)
cc -o edit $(objects)

main.o : defs.h
kbd.o : defs.h command.h
command.o : defs.h command.h
display.o : defs.h buffer.h
insert.o : defs.h buffer.h
search.o : defs.h buffer.h
files.o : defs.h buffer.h command.h
utils.o : defs.h

.PHONY : clean
clean :
-rm edit $(objects)
This is how we would write the makefile in actual practice. (See Section 17.5.7 [Rules for
cleaning the directory], page 163, for the complications associated with clean.)
Because implicit rules are so convenient, they are important. You will see them used
frequently.

17.5.6 Combining rules by prerequisite

When the objects of a makefile are created by implicit rules alone, an alternative style of
makefile is possible. In this style of makefile, you group entries by their prerequisites instead
of by their targets. Here is an example of this alternative style:
objects = main.o kbd.o command.o display.o \
insert.o search.o files.o utils.o

edit : $(objects)
cc -o edit $(objects)

$(objects) : defs.h
kbd.o command.o files.o : command.h
display.o insert.o search.o files.o : buffer.h
Here ‘defs.h’ is given as a prerequisite of all the object files, and ‘command.h’ and
‘buffer.h’ are prerequisites of the specific object files listed for them.
Whether this is better is a matter of taste: it is more compact, but some people dislike
it because they find it clearer to put all the information about each target in one place.
Building a library 163

17.5.7 Rules for cleaning the directory

Compiling a program is not the only thing you might want to write rules for. Makefiles
commonly do a few other things besides compiling a program: for example, they can often
delete all the object files and executables so that the directory is clean.
Here is how we could write a make rule for cleaning our example editor:
clean:
rm edit $(objects)
In practice, we might want to write the rule in a somewhat more complicated manner
to handle unanticipated situations. For example:
.PHONY : clean
clean :
-rm edit $(objects)
This prevents make from getting confused by an actual file called ‘clean’ and causes it to
continue in spite of errors from rm.
A rule such as this should not be placed at the beginning of the makefile, because we do
not want it to run by default! Thus, in the example makefile, we want the rule for ‘edit’,
which recompiles the editor, to remain the default goal.
Since clean is not a prerequisite of ‘edit’, this rule will not run at all if we give the
command make with no arguments. In order to run the rule, we have to type make clean.

17.6 Building a library

We explored what libraries are and how to use them in a previous chapter. (See Chapter 13
[Libraries], page 75, if you need to refresh your memory.) You may have wondered how
libraries are written in the first place. Is the whole process too complicated for a mortal C
programmer to attempt? Not at all.
Suppose you have a function (or set of functions) that you would like to use widely across
the various C programs you write. You might even like to make it available to other users
in a convenient way. To create a code library that will enable you to achieve this, follow
the sequence below. We will use a code example, but you can create your own library by
taking similar steps.
1.
Here’s an example of the kind of function you might like to use in multiple programs.
It accepts one string containing some text to print, and then prints it on the default
printer.
For the sake of example, the file below is named ‘lpr_print.c’.
#include <stdio.h>

void lpr_print (char *the_text)

{
FILE *printer;

printer = popen ("lpr", "w");

fprintf (printer, the_text);
pclose (printer);
}
(See Section 16.4 [Programming with pipes], page 135, for the rationale behind this
function.)
164 Chapter 17: Putting a program together

2.
Now we will create a library.
• To create a static library called ‘liblprprint.a’ containing this function, just
type the following two command lines in your GNU shell:
gcc -c lpr_print.c
ar rs liblprprint.a lpr_print.o
The ‘-c’ option to gcc produces only a ‘.o’ object code file, without linking it,
while the ar command (with its ‘rs’ options) permits the creation of an archive
file, which can contain a bundle of other files that can be re-extracted later (for
example, when executing library code). In this case, we are only archiving one
object code file, but in some cases, you might want to archive multiple ones. (See
the man page for ar for more information.)
• To create a shared library called ‘liblprprint.so’ instead, enter the following
sequence of commands:1
gcc -c -fpic lpr_print.c
gcc -shared -o liblprprint.so lpr_print.o
(For the record, ‘pic’ stands for “position-independent code”, an object-code for-
mat required for shared libraries. You might need to use the option ‘-fPIC’ instead
of ‘-fpic’ if your library is very large.)
3. Now create a header file that will allow users access to the functions in your library.
You should provide one function prototype for each function in your library. Here is a
header file for the library we have created, called ‘liblprprint.h’.
/*
liblprprint.h:
routines in liblprprint.a
and liblprprint.so
*/

extern void lpr_print (char *the_text);

4.
Now you should put your libraries and include file somewhere your code can access
them. For the sake of this example, create the directories ‘include’ and ‘lib’ in your
home directory. Once you have done so, move the ‘.a’ and ‘.so’ files you have created
to ‘lib’, and the ‘.h’ file to ‘include’.
5.
If you have taken the last step, and you want to run a program linked to a shared
version of your library, you should type a line like the following into your shell (the
following command line assumes you are using the Bash shell and that your home
directory is named ‘/home/fred’):
export LD_LIBRARY_PATH=/home/fred/lib:$LD_LIBRARY_PATH
This command line sets an environment variable that makes the linker search the
‘/home/fred/lib’ directory before it searches anywhere else. You can include it in
1
To create library files containing multiple object files, simply include the object files on the same command
line. For example, to create a static library with multiple object files, type a command such as ar rs
liblprprint.a lpr_print.o lpr_print2.o lpr_print3.o. Similarly, to create a shared library, type
gcc -shared -o liblprprint.so lpr_print.o lpr_print2.o lpr_print3.o.
Questions 165

your ‘.bashrc’ or ‘.bash_profile’ file. If you don’t execute this command before you
attempt to run a program using your shared library, you will probably receive an error.
6.
Now you can write programs that use your library. Consider the following short pro-
gram, called ‘printer.c’:
#include <liblprprint.h>

/* To shorten example, not using argp */

int main ()
{
lpr_print ("Hello, Multiverse!\nHowarya?\n");
return 0;
}

To compile this program using your static library, type something like the following
command line:
gcc --static -I../include -L../lib -o printer printer.c -llprprint
The ‘--static’ option forces your static library to be linked; the default is your
shared version. The ‘-llprprint’ option makes GCC link in the ‘liblprprint’
library, just as you would need to type ‘-lm’ to link in the ‘libm’ math library.
The ‘-I../include’ and ‘-L../lib’ options specify that the compiler should look
in the ‘../include’ directory for include files and in the ‘../lib’ directory for
library files. This assumes that you have created the ‘include’ and ‘lib’ directories
in your home directory as outlined above, and that you are compiling your code
in a subdirectory of your home directory. If you are working two directories down,
you would specify ‘-I../../include’, and so on.
The above command line assumes you are using only one ‘.c’ source code file; if
you are using more than one, simply include them on the command line as well.
(See Section 17.4 [Compiling multiple files], page 157.)
Note: Using the ‘--static’ option will force the compiler to link all libraries you
are using statically. If you want to use the static version of your library, but some
shared versions of other libraries, you can omit the ‘--static’ option from the
command line and specify the static version of your library explicitly, as follows:
gcc -I../include -L../lib -o printer printer.c ../lib/liblprprint.a
• To compile this program using your shared library, type something like the follow-
ing command line.
gcc -I../include -L../lib -o printer printer.c -llprprint
7. The executable produced is called ‘printer’. Try it!

17.7 Questions
1. What is the name of the preferred method for handling command-line options?
2. What does the ‘-c’ option of the gcc command do?
3. What information does the argc variable contain?
4. What information does the argv variable contain?
5. What information does the envp variable contain?
166 Chapter 17: Putting a program together
Hidden assignments 167

18 Advanced operators
Concise expressions
In this chapter, we will examine some advanced mathematical and logical operators in
C.

18.1 Hidden operators and values

Many operators in C are more versatile than they appear to be at first glance. Take, for
example, the following operators:
• =
• ++
• --
• +=
• -=
These operators can be used in some surprising ways to make C source code elegant
and compact. (See Chapter 7 [Expressions and operators], page 29, if you need a refresher
in what they do.) All of them can form expressions that have their own values. Such an
expression can be taken as a whole (a “black box”) and treated as a single value, which
can then be assigned and compared to other expressions, in effect, “hidden” within another
expression.
The value of an expression is the result of the operation carried out in the expression.
Increment and decrement statements have a value that is one greater than or one less than
the value of the variable they act upon, respectively.
Consider the following two statements:
c = 5;
c++;
The expression c++ in the above context has the value 6.
Now consider these statements:
c = 5;
c--;
The expression c-- in the above context has the value 4.

18.1.1 Hidden assignments

Assignment expressions have values too — their values are the value of the assignment. For
example, the value of the expression c = 5 is 5.
The fact that assignment statements have values can be used to make C code more
elegant. An assignment expression can itself be assigned to a variable. For example, the
expression c = 0 can be assigned to the variable b:
b = (c = 0);
or simply:
b = c = 0;
These equivalent statements set b and c to the value 0, provided b and c are of the same
type. They are equivalent to the more usual:
168 Chapter 18: Advanced operators

b = 0;
c = 0;
Note: Don’t confuse this technique with a logical test for equality. In the above example,
both b and c are set to 0. Consider the following, superficially similar, test for equality,
however:
b = (c == 0);
In this case, b will only be assigned a zero value (FALSE) if c does not equal 0. If c does
equal 0, then b will be assigned a non-zero value for TRUE, probably 1. (See Section 7.8
[Comparisons and logic], page 34, for more information.)
Any number of these assignments can be strung together:
a = (b = (c = (d = (e = 5))));
or simply:
a = b = c = d = e = 5;
This elegant syntax compresses five lines of code into a single line.
There are other uses for treating assignment expressions as values. Thanks to C’s flexible
syntax, they can be used anywhere a value can be used. Consider how an assignment
expression might be used as a parameter to a function. The following statement gets a
character from standard input and passes it to a function called process_character.
process_character (input_char = getchar());
This is a perfectly valid statement in C, because the hidden assignment statements passes
the value it assigns on to process_character. The assignment is carried out first and then
the process_character function is called, so this is merely a more compact way of writing
the following statements.
input_char = getchar();
process_character (input_char);
All the same remarks apply about the specialized assignment operators +=, *=, /=, and
so on.
The following example makes use of a hidden assignment in a while loop to print out
all values from 0.2 to 20.0 in steps of 0.2.
#include <stdio.h>

/* To shorten example, not using argp */

int main ()
{
double my_dbl = 0;

while ((my_dbl += 0.2) < 20.0)

printf ("%lf ", my_dbl);

printf ("\n");

return 0;
}

18.1.2 Postfix and prefix ++ and --

Increment (++) and decrement (--) expressions also have values, and like assignment ex-
pressions, can be hidden away in inconspicuous places. These two operators are slightly
Arrays and hidden operators 169

more complicated than assignments because they exist in two forms, postfix (for example,
my_var++) and prefix (for example, ++my_var).

Postfix and prefix forms have subtly different meanings. Take the following example:
int my_int = 3;
printf ("%d\n", my_int++);

The increment operator is hidden in the parameter list of the printf call. The variable
my_int has a value before the ++ operator acts on it (3) and afterwards (4).

Which value is passed to printf? Is my_int incremented before or after the printf
call? This is where the two forms of the operator (postfix and prefix) come into play.

If the increment or decrement operator is used as a prefix, the operation is performed

before the function call. If the operator is used as a postfix, the operation is performed
after the function call.

In the example above, then, the value passed to printf is 3, and when the printf
function returns, the value of my_int is incremented to 4. The alternative is to write
int my_int = 3;
printf ("%d\n", ++my_int);

in which case the value 4 is passed to printf.

The same remarks apply to the decrement operator as to the increment operator.

18.1.3 Arrays and hidden operators

Hidden operators can simplify dealing with arrays and strings quite a bit. Hiding operators
inside array subscripts or hidding assignments inside loops can often streamline tasks such as
array initialization. Consider the following example of a one-dimensional array of integers.
#include <stdio.h>

#define ARRAY_SIZE 20

/* To shorten example, not using argp */

int main ()
{
int idx, array[ARRAY_SIZE];

for (idx = 0; idx < ARRAY_SIZE; array[idx++] = 0)

;

return 0;
}

This is a convenient way to initialize an array to zero. Notice that the body of the loop is
completely empty!

Strings can benefit from hidden operators as well. If the standard library function
strlen, which finds the length of a string, were not available, it would be easy to write it
with hidden operators:
170 Chapter 18: Advanced operators

#include <stdio.h>

int my_strlen (char *my_string)

{
char *ptr;
int count = 0;

for (ptr = my_string; *(ptr++) != ’\0’; count++)

;

return (count);
}

/* To shorten example, not using argp */

int main (int argc, char *argv[], char *envp[])
{
char string_ex[] = "Fabulous!";

printf ("String = ’%s’\n", string_ex);

printf ("Length = %d\n", my_strlen (string_ex));

return 0;
}
The my_strlen function increments count while the end of string marker ‘\0’ is not found.
Again, notice that the body of the loop in this function is completely empty.

18.1.4 A warning about style

Overuse of “hidden” operators can produce code that is difficult to understand. See Sec-
tion 22.6 [Hidden operators and style], page 205, for some cautions about when not to use
them.

18.2 The comma operator

The comma operator (,) works almost like the semicolon ‘;’ that separates one C statement
from another. You can separate almost any kind of C statment from another with a comma
operator. The comma-separated expressions are evaluated from left to right and the value
of the whole comma-separated sequence is the value of the rightmost expression in the
sequence. Consider the following code example.
#include <stdio.h>

/* To shorten example, not using argp */

int main (int argc, char *argv[], char *envp[])
{
int a, b, c, d;

a = (b = 2, c = 3, d = 4);
printf ("a=%d\nb=%d\nc=%d\nd=%d\n",
a, b, c, d);
return 0;
}
The value of (b = 2, c = 3, d = 4) is 4 because the value of its rightmost sub-expression,
d = 4, is 4. The value of a is thus also 4. When run, this example prints out the following
text:
Bitwise operators 171

a=4
b=2
c=3
d=4
The comma operator is very useful in for loops. (See Section 11.4 [The flexibility of
for], page 62, for an example.)

18.3 Machine-level operators

Bits and Bytes. Flags. Shifting.
Bits (or binary digits), the values 0 and 1, are the lowest-level software objects in a
computer; there is nothing more primitive. C gives programmers full access to bits and bit
sequences, and even provides high-level operators for manipulating them.
All computer data whatsoever is composed of bit strings. The word “string” is being
used here in its more general sense of sequence; do not confuse the usage with “text string”.
Although all text strings are bit strings, not all bit strings are text strings.
The only difference between a text string and a floating-point value is the way we in-
terpret the pattern of bits in the computer’s memory. For the most part, we can simply
ignore the low level of the computer in which computer data appears as bit strings. Systems
programmers, on the other hand, such as those who wrote GNU/Linux, must frequently
manipulate bits directly in order to handle flags.
A flag is a bit in a bit string that can be either set (1) or cleared (0). We have run across
a few flags already, such as the various flags passed to the GNU C Library functions open;
the flags O_RDONLY and O_WRONLY are actually macros that specify binary values, which can
be manipulated and examined with binary OR and similar functions. Flags are normally
declared as integers in C.
Programmers who perform bit operations on a regular basis often use either octal (base-
8) or hexadecimal (base-16) numbers, because every octal digit specifies exactly three bits,
and every hexadecimal digit specifies four.

18.3.1 Bitwise operators

C provides the following operators for handling bit patterns:
<< Bit-shift left by a specified number of bit positions
>> Bit-shift right by a specified number of bit positions
| Bitwise inclusive OR
^ Bitwise exclusive OR
& Bitwise AND
~ Bitwise NOT
<<= Bit-shift left assignment (var = var << value)
>>= Bit-shift right assignment (var = var >> value)
|= Exclusive OR assignment (var = var | value)
^= Inclusive OR assignment (var = var ^ value)
172 Chapter 18: Advanced operators

&= AND assignment (var = var & value)

The meaning and syntax of these operators is given below.
Don’t confuse bitwise operators (such as bitwise AND, &) with logical operators (such
as logical AND, &&). Bitwise operators operate on each bit in the operand individually.

18.3.2 Shift operations

Imagine a bit string as represented by the following group of boxes. Every box represents
a bit, a binary digit; the ones and zeros inside represent their values. The values written
across the top are the place-values of each bit. (Just as a decimal (base-10) number has a
ones place, a tens place, a hundreds place, a thousands place, and so on, a binary (base-2)
number has the places 1, 2, 4, 8, 16, 32, etc.) The number after the equals sign shows the
value of the bit string in decimal notation.
128 64 32 16 8 4 2 1
-------------------------------
| 0 | 0 | 0 | 0 | 0 | 0 | 0 | 1 | = 1
-------------------------------
Bit-shift operators move whole bit strings left or right. The syntax of the bit-shift left
operation is value << positions ; that of bit-shift right is value >> positions ; So for
example, using the bit string (1) above, the value of 1 << 1 is 2, because the bit string
would have been moved one position to the left:
128 64 32 16 8 4 2 1
-------------------------------
| 0 | 0 | 0 | 0 | 0 | 0 | 1 | 0 | = 2
-------------------------------
Notice how the space to the right of the shifted bit string is simply filled with a 0.
Similarly, the value of 1 << 4 is 16, because the original bit string is shifted left four
places:
128 64 32 16 8 4 2 1
-------------------------------
| 0 | 0 | 0 | 1 | 0 | 0 | 0 | 0 | = 16
-------------------------------
Notice, again, that the spaces to the right of the original bit string are filled out with zeros.
Now for a slightly more difficult one. The value of 6 << 2 is 24. Here is the bit string
representing 6:
128 64 32 16 8 4 2 1
-------------------------------
| 0 | 0 | 0 | 0 | 0 | 1 | 1 | 0 | = 6
-------------------------------
Shift 6 left 2 places:
128 64 32 16 8 4 2 1
-------------------------------
| 0 | 0 | 0 | 1 | 1 | 0 | 0 | 0 | = 24
-------------------------------
Notice that every shift left multiplies by 2. (Since 6 << 2 means to shift 6 left twice, the
result is 24.)
As you might expect, every shift right performs (integer) division by two on the number.
If a bit is shifted beyond the ones position (the rightmost “box”), however, then it “drops
off” and is lost. So the following equalities hold:
Bitwise exclusive OR (XOR/EOR) 173

1 >> 1 == 0
2 >> 1 == 1
2 >> 2 == 0
n >> n == 0

One common use of bit-shifting is to scan through the bits of a bit-string one by one in
a loop. This is done with bit masks, as described in the next section.

18.3.3 Truth tables and bit masks

The binary operators AND (&), OR (inclusive OR, |) and XOR (exclusive OR, also called
EOR, ^) perform comparisons, or masking operations, between two bit strings. They are
also binary operators in the sense that they take two operands. There is another operator
called NOT (~) that is a unary operator; it takes only one operand.
These bitwise operations are best summarized by truth tables. Each truth table for a
binary operator (that is, one with two operands), indicates what the result of the operation
is for every possible combination of two bits.

18.3.3.1 Bitwise NOT

The unary operator NOT (~) simply generates the one’s complement of the bit string; that
is, it returns the same bit string, with all ones replaced with zeros and all zeros replaced
with ones. As a truth table this would be summarized as follows:
~value result
0 1
1 0

18.3.3.2 Bitwise AND

Bitwise AND operates on two values, for example 0 & 1. Both the first value and the second
value must be 1 in order for the result to be 1. As a truth table this would be summarized
as follows:
value1 value2 result
0 0 0
0 1 0
1 0 0
1 1 1

18.3.3.3 Bitwise inclusive OR

Bitwise OR operates on two values, for example 0 | 1. The result is 1 if the first value or
the second value is 1, or both are 1. As a truth table this would be summarized as follows:
value1 value2 result
0 0 0
0 1 1
1 0 1
1 1 1
174 Chapter 18: Advanced operators

18.3.3.4 Bitwise exclusive OR (XOR/EOR)

Bitwise XOR operates on two values, for example 0 ^ 1. The result is 1 if the first value or
the second value is 1, but not if both are 1 (hence the name “exclusive OR”). As a truth
table this would be summarized as follows:
value1 value2 result
0 0 0
0 1 1
1 0 1
1 1 0

18.3.3.5 Masks
Bit strings and bitwise operators are often used to make masks. A mask is a bit string that
“fits over” another bit string and produces a desired result, such as singling out particular
bits from the second bit string, when the two bit strings are operated upon. This is par-
ticularly useful for handling flags; programmers often wish to know whether one particular
flag is set in a bit string, but may not care about the others. For example, you might create
a mask that only allows the flag of interest to have a non-zero value, then AND that mask
with the bit string containing the flag.
Consider the following mask, and two bit strings from which we want to extract the final
bit:
mask = 00000001
value1 = 10011011
value2 = 10011100

mask & value1 == 00000001

mask & value2 == 00000000
The zeros in the mask mask off the first seven bits and only let the last bit show through.
(In the case of the first value, the last bit is 1; in the case of the second value, the last bit
is 0.)
Alternatively, masks can be built up by operating on several flags, usually with inclusive
OR:
flag1 = 00000001
flag2 = 00000010
flag3 = 00000100

mask = flag1 | flag2 | flag3

mask == 00000111
See Section 16.5.2 [Opening files at a low level], page 138, for a code example that
actually uses bitwise OR to join together several flags.
It should be emphasized that the flag and mask examples are written in pseudo-code, that
is, a means of expressing information that resembles source code, but cannot be compiled.
It is not possible to use binary numbers directly in C.
The following code example shows how bit masks and bit-shifts can be combined. It
accepts a decimal number from the user between 0 and 128, and prints out a binary number
in response.
Questions 18 175

#include <stdio.h>
#define NUM_OF_BITS 8

/* To shorten example, not using argp */

int main ()
{
char *my_string;
int input_int, args_assigned;
int nbytes = 100;
short my_short, bit;
int idx;

/* This hex number is the same as binary 10000000 */

short MASK = 0x80;

args_assigned = 0;
input_int = -1;

while ((args_assigned != 1) ||
(input_int < 0) || (input_int > 128))
{
puts ("Please enter an integer from 0 to 128.");
my_string = (char *) malloc (nbytes + 1);
getline (&my_string, &nbytes, stdin);
args_assigned = sscanf (my_string, "%d", &input_int);
if ((args_assigned != 1) ||
(input_int < 0) || (input_int > 128))
puts ("\nInput invalid!");
}

my_short = (short) input_int;

printf ("Binary value = ");

/*
Convert decimal numbers into binary
Keep shifting my_short by one to the left
and test the highest bit. This does
NOT preserve the value of my_short!
*/

for (idx = 0; idx < NUM_OF_BITS; idx++)

{
bit = my_short & MASK;
printf ("%d", bit/MASK);
my_short <<= 1;
}

printf ("\n");
return 0;
}

18.4 Questions 18
1. Hidden operators can be used in return statements, for example,
return (++x);
Would there be any point in writing the following?
176 Chapter 18: Advanced operators

return (x++);
2. What distinguishes a bit string from an ordinary variable? Can any variable be a bit
string?
3. What is the difference between an inclusive OR operation and an exclusive OR opera-
tion?
4. Find out what the decimal values of the following operations are.
1. 7 & 2
2. 1 & 1
3. 15 & 3
4. 15 & 7
5. 15 & 7 & 3
Try to explain the results. (Hint: sketch out the numbers as bit strings.)
5. Find out what the decimal values of the following operations are.
1. 1 | 2
2. 1 | 2 | 3
6. Find out the decimal values of the following operations.
1. 1 & (~1)
2. 23 & (~23)
3. 2012 & (~2012)
(Hint: write a short program to work them out.)
enum 177

19 More data types

There are still a few data types in C that we have not discussed. Actually, since C allows
you to define new data types at will, no one can ever cover all possibilities. We will only
discuss the most important examples.
enum Type specifier for variables that can have a set of different values.
void Type specifier for “empty” data.
volatile Type qualifier for data that changes independently of the program.
const Type qualifier for data that cannot change.
In addition, there are two data types called struct and union that are so important,
they have received their own chapter. (See Chapter 20 [Data structures], page 183, for more
information on struct and union.)

19.1 enum
The enum type specifier is short for “enumerated data”. The user can define a fixed set of
words that a variable of type enum can take as its value. The words are assigned integer
values by the compiler so that code can compare enum variables. Consider the following
code example:
#include <stdio.h>

/* To shorten example, not using argp */

int main ()
{
enum compass_direction
{
north,
east,
south,
west
};

enum compass_direction my_direction;

my_direction = west;

return 0;
}
This example defines an enumerated variable type called compass_direction, which can
be assigned one of four enumerated values: north, east, south, or west. It then declares
a variable called my_direction of the enumerated compass_direction type, and assigns
my_direction the value west.
Why go to all this trouble? Because enumerated data types allow the programmer to
forget about any numbers that the computer might need in order to process a list of words,
and simply concentrate on using the words themselves. It’s a higher-level way of doing
things; in fact, at a lower level, the computer assigns each possible value in an enumerated
data type an integer cconstant — one that you do not need to worry about.
Enumerated variables have a natural partner in the switch statement, as in the following
code example.
178 Chapter 19: More data types

#include <stdio.h>

enum compass_direction
{
north,
east,
south,
west
};

enum compass_direction get_direction()

{
return south;
}

/* To shorten example, not using argp */

int main ()
{
enum compass_direction my_direction;
puts ("Which way are you going?");
my_direction = get_direction();

switch (my_direction)
{
case north:
puts("North? Say hello to the polar bears!");
break;

case south:
puts("South? Say hello to Tux the penguin!");
break;

case east:
puts("If you go far enough east, you’ll be west!");
break;

case west:
puts("If you go far enough west, you’ll be east!");
break;
}

return 0;
}

In this example, the compass_direction type has been made global, so that the get_
direction function can return that type. The main function prompts the user, ‘Which way
are you going?’, then calls the “dummy” function get_direction. In a “real” program,
such a function would accept input from the user and return an enumerated value to main,
but in this case it merely returns the value south. The output from this code example is
therefore as follows:
Which way are you going?
South? Say hello to Tux the penguin!

As mentioned above, enumerated values are converted into integer values internally by
the compiler. It is practically never necessary to know what integer values the compiler
assigns to the enumerated words in the list, but it may be useful to know the order of the
void 179

enumerated items with respect to one another. The following code example demonstrates
this.
#include <stdio.h>

/* To shorten example, not using argp */

int main ()
{
enum planets
{
Mercury,
Venus,
Earth,
Mars,
Jupiter,
Saturn,
Uranus,
Neptune,
Pluto
};

enum planets planet1, planet2;

planet1 = Mars;
planet2 = Earth;

if (planet1 > planet2)

puts ("Mars is farther from the Sun than Earth is.");
else
puts ("Earth is farther from the Sun than Mars is.");

return 0;
}
The output from this example reads as follows:
Mars is farther from the Sun than Earth is.

19.2 void
The void data type was introduced to make C syntactically consistent. The main reason
for void is to declare functions that have no return value. The word “void” is therefore
used in the sense of “empty” rather than that of “invalid”.
C functions are considered by the compiler to return type int unless otherwise specified.
Although the data returned by a function can legally be ignored by the function calling it,
the void data type was introduced by the ANSI standard so that C compilers can issue
warnings when an integer value is not returned by a function that is supposed to return
one. If you want to write a function that does not return a value, simply declare it void. A
function declared void has no return value and simply returns with the command return;.
Variables can be declared void as well as functions:
void my_variable;
void *my_pointer;
A variable that is itself declared void (such as my_variable above) is useless; it cannot
be assigned a value, cannot be cast to another type, in fact, cannot be used in any way.
Void pointers (type void *) are a different case, however. A void pointer is a generic
pointer; any pointer can be cast to a void pointer and back without any loss of information.
180 Chapter 19: More data types

Any type of pointer can be assigned to (or compared with) a void pointer, without casting
the pointer explicitly.
Finally, a function call can be cast to void in order to explicitly discard a return value.
For example, printf returns a value, but it is seldom used. Nevertheless, the two lines of
code that follow are equivalent:
printf ("Hullo!\n");
(void) printf ("Hullo!\n");
There is no good reason to prefer the second line to the first, however, so using the more
concise form is preferred.

19.3 volatile
The volatile type qualifier was introduced by the ANSI Standard to permit the use of
memory-mapped variables, that is, variables whose value changes autonomously based on
input from hardware. One might declare a volatile variable volatile float temperature;
whose value fluctuated according to readings from a digital thermometer connected to the
computer.
There is another use for the volatile qualifier that has to do with multiprocessing
operating systems. Independent processes that share common memory might each change
the value of a variable independently. The volatile keyword serves as a warning to the
compiler that it should not optimize the code containing the variable (that is, compile it so
that it will run in the most efficient way possible) by storing the value of the variable and
referring to it repeatedly, but should reread the value of the variable every time. (Volatile
variables are also flagged by the compiler as not to be stored in read-only memory.)

19.4 Constants
Constants in C usually refer to two things: either a type of variable whose value cannot
change declared with the const qualifier (in this case, “variable” is something of a mis-
nomer), or a string or numeric value incorporated directly into C code, such as ‘1000’. We
will examine both kinds of constant in the next two sections.

19.4.1 const
Sometime a variable must be assigned a value once and once only; for example, it might be
in read-only memory. The reserved word const is, like static and volatile, a data type
qualifier that can be applied to many different data types. It declares a variable to be a
constant, whose value cannot be reassigned. A const must be assigned a value when it is
declared.
const double avogadro = 6.02e23;
const int moon_landing = 1969;
You can also declare constant arrays:
const int my_array[] =
{0, 1, 2, 3, 4, 5, 6, 7, 8};
Any attempt to assign a new value to a const variable will result in a compile-time error
such as the following:
const.c: In function ‘main’:
const.c:11: warning: assignment of read-only variable ‘avogadro’
typedef 181

19.4.2 Constant expressions

You can declare constant expressions explicitly as a particular type of value, such as a long
integer, a float, a character, or a hexadecimal value, with certain typographical conventions.
For example, it is possible to declare a value explicitly as a long by placing the letter ‘L’
after the numeric constant. For example:
#define MY_LONG1 23L;
#define MY_LONG2 236526598L;

Similarly, you can declare a value to be a float by appending the letter ‘F’ to it. Of
course, numeric constants containing a decimal point are automatically considered floats.
The following constants are both floating-point numbers:
#define MY_FLOAT1 23F;
#define MY_FLOAT2 23.5001;

You can declare a hexadecimal (base-16) number by prefixing it with ‘0x’; you can declare
an octal (base-8) number by prefixing it with ‘0’. For example:
int my_hex_integer = 0xFF; /* hex FF */
int my_octal_integer = 077; /* octal 77 */

You can use this sort of notation with strings and character constants too. ASCII
character values range from 0 to 255. You can print any character in this range by prefixing a
hexadecimal value with ‘\x’ or an octal value with ‘\’. Consider the following code example,
which demonstrates how to print the letter ‘A’, using either a hexadecimal character code
(‘\x41’) or an octal one (‘\101’).
#include <stdio.h>

/* To shorten example, not using argp */

int main ()
{
printf ("\\x41 hex = \x41\n");
printf ("\\101 octal = \101\n");

return 0;
}

The preceding code prints the following text:

\x41 hex = A
\101 octal = A

Of course, you can assign a variable declared with the const qualifier (the first kind
of “constant” we examined) a constant expression declared with one of the typographical
expressions above. For example:
const int my_hex_integer = 0xFF; /* hex FF */
const int my_octal_integer = 077; /* octal 77 */

19.5 struct and union

Structures and unions are data types that are important enough to merit a chapter of their
own. See Chapter 20 [Data structures], page 183, for more information on structures and
unions.
182 Chapter 19: More data types

19.6 typedef
You can define your own data types in C with the typedef command, which may be written
inside functions or in global scope. This statement is used as follows:
typedef existing_type new_type ;
You can then use the new type to declare variables, as in the following code example,
which declares a new type called my_type and declares three variables to be of that type.
#include <stdio.h>

/* To shorten example, not using argp */

int main (int argc, char *argv[], char *envp[])
{
typedef int my_type;
my_type var1, var2, var3;

var1 = 10;
var2 = 20;
var3 = 30;

return 0;
}
The new type called my_type behaves just like an integer. Why, then, would we use it
instead of integer?
Actually, you will seldom wish to rename an existing data type. The most important
use for typedef is in renaming structures and unions, whose names can become long and
tedious to declare otherwise. We’ll investigate structures and unions in the next chapter.
(See Chapter 20 [Data structures], page 183.)

19.7 Questions 19
1. Enumerated names are given integer values by the compiler so that it can do multipli-
cation and division with them. True or false?
2. Does void do anything which C cannot already do without this type?
3. What type qualifier might a variable accessed directly by a timer be given?
4. Write a statement which declares a new type "real" to be like the usual type "double".
5. Variables declared with the qualifier const can be of any type. True or false?
Structure declarations 183

20 Data structures
Grouping data. Tidying up programs.
It would be hard for a program to manipulate data if it were scattered around with no
particular structure. C therefore has several facilities to group data together in convenient
packages, or data structures. One type of data structure in C is the struct (or structure)
data type, which is a group of variables clustered together with a common name. A related
data type is the union, which can contain any type of variable, but only one at a time.
Finally, structures and unions can be linked together into complex data structures such as
lists and trees. This chapter explores all of these kinds of data structure.
It is important to distinguish the terms structure and data structure. “Data structure”
is a generic term that refers to any pattern of data in a computer program. An array is a
data structure, as is a string. A structure is a particular data type in C, the struct; all
struct variables (structures) are data structures, but not all data structures are structures.

20.1 struct
A structure is a group of one or more variables under a single name. Unlike arrays, structures
can contain a combination of different types of data; they can even contain arrays. A
structure can be arbitrarily complex.
Every type of structure that is defined is given a name, and the variables it contains
(called members) are also given names. Finally, every variable declared to be of a particular
structure type has its own name as well, just as any other variable does.

20.1.1 Structure declarations

The following statement is a type declaration, so it belongs with other declarations, either
at the start of a program or the start of a code block.
struct personal_data
{
char name[100];
char address[200];
int year_of_birth;
int month_of_birth;
int day_of_birth;
};
The statement says: define a type of variable that holds a string of 100 characters called
name, a string of 200 characters called address, and three integers called year_of_birth,
month_of_birth, and day_of_birth. Any variable declared to be of type struct
personal_data will contain these components, which are called members. Different
structures, even different types of structure, can have members with the same name, but
the values of members of different structures are independent of one another. You can also
use the same name for a member as for an ordinary variable in your program, but the
computer will recognize them as different entities, with different values. This is similar to
the naming convention for humans, where two different men may share the name "John
Smith", but are recognized as being different people.
Once you have declared a type of structure, you can declare variables to be of that type.
For example:
184 Chapter 20: Data structures

struct personal_data person0001;

The statement above declares a variable called person0001 to be of type struct
personal_data. This is probably the most common method of declaring a structure
variable, but there are two equivalent methods. For example, a structure variable can be
declared immediately after the declaration of the structure type:
struct personal_data
{
char name[100];
char address[200];
int year_of_birth;
int month_of_birth;
int day_of_birth;
} person0001;

20.1.1.1 Structure declarations using typedef

Alternatively, the typedef command can be used to cut down on typing out code in the
long term. The type definition is made once at the start of the program and subsequent
variable declarations are made by using the new name, without the word struct:
typedef struct
{
char name[100];
char address[200];
int year_of_birth;
int month_of_birth;
int day_of_birth;
} personal_data;

personal_data person001;
personal_data person002;
personal_data person003;
Note that this use of the typedef command parallels the usage we have already seen:
typedef existing_type new_type
In the example above of using typedef to declare a new type of structure, the metasyn-
tactic variable new type corresponds to the identifier personal_data, and the metasyntac-
tic variable existing type corresponds to the following code:
struct
{
char name[100];
char address[200];
int year_of_birth;
int month_of_birth;
int day_of_birth;
}
Structure type and variable declarations can be either local or global, depending on their
placement in the code, just as any other declaration can be.

20.1.2 Using structures

Structures are extremely powerful data types. Not only can you pass a whole structure as
a parameter to a function, or return one as a value from a function. You can even assign
one structure to another.
Arrays of structures 185

You can get and set the values of the members of a structure with the ‘.’ dot character.
This is called the member operator. The general form of a member reference is:
structure_name.member_name
In the following example, the year 1852 is assigned to the year_of_birth member of the
structure variable person1, of type struct personal_data. Similarly, month 5 is assigned
to the month_of_birth member, and day 4 is assigned to the day_of_birth member.
struct personal_data person1;

person1.year_of_birth = 1852;
person1.month_of_birth = 5;
person1.day_of_birth = 4;
Besides the dot operator, C also provides a special -> member operator for use in con-
junction with pointers, because pointers and structures are used together so often. (See
Section 20.1.5 [Pointers to structures], page 187.)
Structures are easy to use For example, you can assign one structure to another structure
of the same type (unlike strings, for example, which must use the string library routine
strcpy). Here is an example of assigning one structure to another:
struct personal_data person1, person2;

person2 = person1;
The members of the person2 variable now contain all the data of the members of the
person1 variable.
Structures are passed as parameters in the usual way:
my_structure_fn (person2);
You would declare such a function thus:
void my_structure_fn (struct personal_data some_struct)
{
}
Note that in order to declare this function, the struct personal_data type must be de-
clared globally.
Finally, a function that returns a structure variable would be declared thusly:
struct personal_data structure_returning_fn ()
{
struct personal_data random_person;
return random_person;
}
Of course, random_person is a good name for the variable returned by this bare-bones
function, because without unless one writes code to initialize it, it can only be filled with
garbage values.

20.1.3 Arrays of structures

Just as arrays of basic types such as integers and floats are allowed in C, so are arrays of
structures. An array of structures is declared in the usual way:
struct personal_data my_struct_array[100];
The members of the structures in the array are then accessed by statements such as the
following:
186 Chapter 20: Data structures

The value of a member of a structure in an array can be assigned to another variable, or the
value of a variable can be assigned to a member. For example, the following code assigns the
number 1965 to the year_of_birth member of the fourth element of my_struct_array:
my_struct_array[3].year_of_birth = 1965;
(Like all other arrays in C, struct arrays start their numbering at zero.)
The following code assigns the value of the year_of_birth member of the fourth element
of my_struct_array to the variable yob:
yob = my_struct_array[3].year_of_birth;
Finally, the following example assigns the values of all the members of the second element
of my_struct_array, namely my_struct_array[1], to the third element, so my_struct_
array[2] takes the overall value of my_struct_array[1].
my_struct_array[2] = my_struct_array[1];

20.1.4 Nested structures

Structures can contain other structures as members; in other words, structures can nest.
Consider the following two structure types:
struct first_structure_type
{
int integer_member;
float float_member;
};

struct second_structure_type
{
double double_member;
struct first_structure_type struct_member;
};
The first structure type is incorporated as a member of the second structure type. You can
initialize a variable of the second type as follows:
struct second_structure_type demo;

demo.double_member = 12345.6789;
demo.struct_member.integer_member = 5;
demo.struct_member.float_member = 1023.17;
The member operator ‘.’ is used to access members of structures that are themselves mem-
bers of a larger structure. No parentheses are needed to force a special order of evaluation;
a member operator expression is simply evaluated from left to right.
In principle, structures can be nested indefinitely. Statements such as the following are
syntactically acceptable, but bad style. (See Chapter 22 [Style], page 203.)
my_structure.member1.member2.member3.member4 = 5;
What happens if a structure contains an instance of its own type, however? For example:
struct regression
{
int int_member;
struct regression self_member;
};
In order to compile a statement of this type, your computer would theoretically need an
infinite amount of memory. In practice, however, you will simply receive an error message
along the following lines:
Initializing structures 187

struct5.c: In function ‘main’:

struct5.c:8: field ‘self_member’ has incomplete type

The compiler is telling you that self_member has been declared before its data type,
regression has been fully declared — naturally, since you’re declaring self_member in the
middle of declaring its own data type!

20.1.5 Pointers to structures

Although a structure cannot contain an instance of its own type, it can can contain a
pointer to another structure of its own type, or even to itself. This is because a pointer
to a structure is not itself a structure, but merely a variable that holds the address of a
structure. Pointers to structures are quite invaluable, in fact, for building data structures
such as linked lists and trees. (See Section 20.4 [Complex data structures], page 192.)

A pointer to a structure type variable is declared by a statement such as the following:

struct personal_data *my_struct_ptr;

The variable my_struct_ptr is a pointer to a variable of type struct personal_data. This

pointer can be assigned to any other pointer of the same type, and can be used to access
the members of its structure. According to the rules we have outlined so far, this would
have to be done like so:

struct personal_data person1;

my_struct_ptr = &person1;
(*my_struct_ptr).day_of_birth = 23;

This code example says, in effect, “Let the member day_of_birth of the structure pointed
to by my_struct_ptr take the value 23.” Notice the use of parentheses to avoid confusion
about the precedence of the ‘*’ and ‘.’ operators.

There is a better way to write the above code, however, using a new operator: ‘->’. This
is an arrow made out of a minus sign and a greater than symbol, and it is used as follows:

my_struct_ptr->day_of_birth = 23;

The ‘->’ enables you to access the members of a structure directly via its pointer. This
statement means the same as the last line of the previous code example, but is consider-
ably clearer. The ‘->’ operator will come in very handy when manipulating complex data
structures. (See Section 20.4 [Complex data structures], page 192.)

20.1.6 Initializing structures

In the chapter on arrays, we explored how to initialize an array with values at compile time.
(See Section 14.5 [Initializing arrays], page 92.) It is also possible to initialize structures at
compile time, as shown below. (This code example also shows how to dynamically allocate
structures with malloc and initialize them with the -> operator. See Section 20.2 [Memory
allocation], page 189, for more information on this technique.)
188 Chapter 20: Data structures

#include <stdio.h>

/* To shorten example, not using argp */

int main()
{
struct personal_data
{
char name[100];
char address[200];
int year_of_birth;
int month_of_birth;
int day_of_birth;
};

struct personal_data person1 =

{
"Liddell, Alice",
"Wonderland",
1852,
5,
4
};

struct personal_data person2 =

{
"Hale-Evans, Ron",
"Seattle, Washington",
1965,
6,
27
};

struct personal_data* person_ptr1;

struct personal_data* person_ptr2;

person_ptr1 = (struct personal_data*)

malloc (sizeof (struct personal_data));

strcpy (person_ptr1->name, "Adams, Douglas");

strcpy (person_ptr1->address, "The Galaxy");
person_ptr1->year_of_birth = 1952;
person_ptr1->month_of_birth = 3;
/* Don’t know his exact birthday */

person_ptr2 = (struct personal_data*)

malloc (sizeof (struct personal_data));

strcpy (person_ptr2->name, "Egan, Greg");

strcpy (person_ptr2->address, "Permutation City");
person_ptr2->year_of_birth = 1961;
/* Don’t know his birthday */
Memory allocation 189

puts ("Data contained:");

puts (person1.name);
puts (person2.name);
puts (person_ptr1->name);
puts (person_ptr2->name);

return 0;
}
Any trailing items not initialized by data you specify are set to zero.

20.2 Memory allocation

Most variables in C have a fixed size. For example, a string declared to be 200 bytes long
will always be 200 bytes long throughout the program. Sometimes, however, you will need
variables whose size can vary. You might want a string whose size can vary between 0
and 100 kilobytes, for instance. We have already seen occasions where this sort of string is
needed with the getline function. (See Section 16.2.6.1 [getline], page 120.)
This is where dynamic data, or data whose size can vary, comes in. Dynamic data is
created via the process of memory allocation, that is, assigning a block of memory to a vari-
able. Blocks of memory are usually assigned with the malloc function (the function name
is from the phrase “memory allocation”), and can be resized with the realloc (“memory
reallocation”) function, and even merged back into the pool of available memory with the
free function.
The malloc function takes one argument, the number of bytes to allocate. It returns a
void pointer, which provides the address of the beginning of a block of memory that the
program can use. This void pointer can be assigned to any other type of pointer. The only
way to make use of the block of memory that has been allocated is through its pointer; in
that sense, the block is not a “real” variable, that is to say, you cannot assign a value to
the memory block directly. Instead, the address returned by malloc enables you to use the
block indirectly; in this way, the block can contain any kind of value a real variable can.
Having to use blocks indirectly through pointers is a small price to pay for the flexibility of
dynamic data.
The following code example allocates a ten-byte string:
char *my_string;
my_string = (char *) malloc(10+1);
Notice that the void pointer returned by malloc is cast to a character pointer (type char
*) before it is assigned to my_string. (See Section 5.4 [The cast operator], page 20.) Also
notice that we have actually allocated 11 bytes of space; this is because the 11th byte must
contain a null character that terminates the string but does not count toward its actual
length. Careful! The newly-allocated block will be filled with garbage.
To reallocate the memory, use the realloc function. This function takes two parameters.
The first is the pointer to the memory block to be reallocated, and the second is a number
of type size_t that specifies the new size for the block. It returns a void pointer to the
newly reallocated block. Here is how to reallocate the block allocated for my_string above,
to a new size of 1000 bytes:
my_string = (char *) realloc (my_string, 1001);
The new block will contain all the data in the old block, followed by enough space to pad
out the block to the new length. The new space will be filled with garbage.
190 Chapter 20: Data structures

Finally, to free up the memory allocated to a block and return it to the common pool of
memory available to your program, use the free function, which takes only one argument,
the pointer to the block you wish to free. It does not return a value.
free (my_string);
It is also possible to allocate the memory for a structure when it is needed and use the
‘->’ operator to access the members of the structure, since we must access the structure
via a pointer. (See the code sample following the next paragraph for an example of how to
do this.) If you are creating complex data structures that require hundreds or thousands of
structure variables (or more), the ability to create and destroy them dynamically can mean
quite a savings in memory.
It’s easy enough to allocate a block of memory when you know you want 1000 bytes for a
string, but how do you know how much memory to allocate for a structure? For this task, C
provides the sizeof function, which calculates the size of an object. For example, sizeof
(int) returns the numbers of bytes occupied by an integer variable. Similarly, sizeof
(struct personal_data) returns the number of bytes occupied by our personal_data
structure. To allocate a pointer to one of these structures, then set the year_of_birth
member to 1852, you would write something like the following:
struct personal_data* my_struct_ptr;

my_struct_ptr = (struct personal_data*)

malloc (sizeof (struct personal_data));
my_struct_ptr->year_of_birth = 1852;

20.3 union
A union is like a structure in which all of the members are stored at the same address. Only
one member can be in a union at one time. The union data type was invented to prevent
the computer from breaking its memory up into many inefficiently sized chunks, a condition
that is called memory fragmentation.
The union data type prevents fragmentation by creating a standard size for certain data.
When the computer allocates memory for a program, it usually does so in one large block
of bytes. Every variable allocated when the program runs occupies a segment of that block.
When a variable is freed, it leaves a “hole” in the block allocated for the program. If this
hole is of an unusual size, the computer may have difficulty allocating another variable to
“fill” the hole, thus leading to inefficient memory usage. Since unions have a standard data
size, however, any “hole” left in memory by freeing a union can be filled by another instance
of the same type of union. A union works because the space allocated for it is the space
taken by its largest member; thus, the small-scale memory inefficiency of allocating space
for the worst case leads to memory efficiency on a larger scale.

20.3.1 Declaration of unions

A union is declared in the same way as a structure. It has a list of members, as in the
example below:
union int_or_float
{
int int_member;
float float_member;
};
Complex data structures 191

Declaring union variables is similar to declaring structure variables:

union int_or_float my_union1, my_union2;

Just like structures, the members of unions can be accessed with the ‘.’ and ‘->’ oper-
ators. However, unlike structures, the variables my_union1 and my_union2 above can be
treated as either integers or floating-point variables at different times during the program.
For example, if you write my_union1.int_member = 5;, then the program sees my_union1
as being an integer. (This is only a manner of speaking. However, my_union1 by itself
does not have a value; only its members have values.) On the other hand, if you then
type my_union1.float_member = 7.7;, the my_union variable loses its integer value. It is
crucial to remember that a union variable can only have one type at the same time.

20.3.2 Using unions

One way to tell what type of member is currently stored in the union is to maintain a flag
variable for each union. This can be done easily with enumerated data. For example, for
the int_or_float type, we might want an associated enumerated type like this:
enum which_member
{
INT,
FLOAT
};

Notice that we used all-uppercase letters for the enumerated values. We would have received
a syntax error if we had actually used the C keywords int and float.
Associated union and enumerated variables can now be declared in pairs:
union int_or_float my_union1;
enum which_member my_union_status1;

Handling union members is now straightforward. For example:

switch (my_union_status1)
{
case INT:
my_union1.int_member += 5;
break;
case FLOAT:
my_union1.float_member += 23.222333;
break;
}

These variables could even be grouped into a structure for ease of use:
struct multitype
{
union int_or_float number;
enum which_member status;
};

struct multitype my_multi;

You would then make assignments to the members of this structure in pairs:
my_multi.number.int_member = 5;
my_multi.status = INT;
192 Chapter 20: Data structures

20.4 Complex data structures

When building data structures, it is best to model the situation in question clearly and
efficiently. Different types of data structure are good for different things. For example,
arrays are good for storing tabular information. A chessboard looks like a two-dimensional
array, so such an array is a good data structure to model a chess game. In this section
we will examine more complex data structures that are useful for modeling more complex
situations.

20.4.1 Data structure diagrams

Sometimes you will want to draw a picture that shows how to solve a problem by displaying
how all its parts are connected. Such a picture is called a structure diagram.
Consider a hypothetical application that stores a map of the local countryside. This
program must store information about individual towns and be able to give directions to
the user about how to get from one town to another. A person driving a car from town
to town might use a map, but the application programmer might use a structure diagram.
Here is a structure diagram for the imaginary town of New Haven, and its neighboring
towns North Haven, East Haven, South Haven, and West Haven:
North Haven

^
|
v

West Haven <-> New Haven <-> East Haven

^
|
v

South Haven
Once you have a structure diagram that represents your information, you can create a
data structure that translates the structure diagram into the computer’s memory. In this
case, we can create a “town structure” that contains pointers to the towns that lie at the
end of roads in the various compass directions. The town structure might look something
like this:
struct town
{
struct town *north;
struct town *south;
struct town *east;
struct town *west;
char name[50];
};
If the user of this hypothetical application wishes to know what is to the north of a
particular town, the program only has to check that town’s north pointer.

20.4.2 Dynamic data structures, Pointers and Dynamic Memory

For programs dealing with large sets of data, it would be a nuisance to have to name every
structure variable containing every piece of data in the code — for one thing, it would be
Binary trees 193

inconvenient to enter new data at run time because you would have to know the name of
the variable in which to store the data when you wrote the program. For another thing,
variables with names are permanent — they cannot be freed and their memory reallocated,
so you might have to allocate an impractically large block of memory for your program at
compile time, even though you might need to store much of the data you entered at run
time temporarily.
Fortunately, complex data structures are built out of dynamically allocated memory,
which does not have these limitations. All your program needs to do is keep track of a
pointer to a dynamically allocated block, and it will always be able to find the block.
A complex data structure is usually built out of the following components:
nodes Dynamically-allocated blocks of data, usually structures.
links Pointers from nodes to their related nodes.
root The node where a data structure starts, also known as the root node. The
address of the root of a data structure must be stored explicitly in a C variable,
or else you will lose track of it.
There are some advantages to the use of dynamic storage for data structures:
• As mentioned above, since memory is allocated as needed, we don’t need to declare
how much we shall use in advance.
• Complex data structures can be made up of lots of “lesser” data structures in a modular
way, making them easier to program.
• Using pointers to connect structures means that they can be re-connected in different
ways as the need arises. (Data structures can be sorted, for example.)

20.4.3 Lists and trees

Two data structures that use nodes and links are very common: the linked list and the
binary tree

20.4.3.1 Linked lists.

A linked list is a linear sequence of structures joined together by pointers. Each node’s
pointer links to the next node in the sequence. Linked lists have two main advantages over
one dimensional arrays: they can be sorted easily simply by redirecting pointers, and they
can be made any length at all dynamically.
Here is an example of a structure type from a linked list:
struct list_node
{
double value;
struct list_node *next;
};
Here the value member holds the actual content of the node, in this case a double-precision
floating-point number, and the next member points to the next node in the list.
You will often encounter another basic kind of linked list, called a doubly-linked list.
Each node in a doubly-linked list contains not only a pointer to the next node, but also to
the previous node. This kind of list makes it easier to determine what the node preceding
a node is, as well as the node succeeding it.
194 Chapter 20: Data structures

20.4.3.2 Binary trees

A binary tree is a data structure in which each node contains links to two successor nodes,
so that the whole structure is shaped like a branching tree. A typical use for a binary tree
might be storing genealogical information; since (at this point in human evolution) every
individual has two parents, each node can represent a person and the two linked nodes can
represent that person’s mother and father. Let’s extend our personal_data structure to
incorporate this kind of information:
struct personal_data
{
char name[100];
char address[200];
int year_of_birth;
int month_of_birth;
int day_of_birth;

struct personal_data *mother;

struct personal_data *father;
};

20.4.4 Setting up a data structure

Plan your data structures well, before you write any program code. Changes in program
code may not affect data structures, but changes to data structures will likely imply drastic
changes to program code.

20.4.4.1 Designing your data structure

The steps you should take in designing a data structure follow a basic pattern:
1. Group together all the kinds of information that must be stored and define a structure
type with a member for each kind of information.
2. Add structure pointers to the structure type to reflect the way in which each bundle
of information is connected to the others.
3. Design the algorithms to handle the memory allocation, node linking, and data storage.

20.4.4.2 Initializing your data structure

Once you understand your data structure, you can set about initializing it in the following
way:
1. Declare your structure type. For example:
struct town
{
struct town *north;
struct town *south;
struct town *east;
struct town *west;
char name[50];
};
2. Declare two pointers to this type:
struct town *root, *current;
The root pointer is used to point to the root node of the data structure, and the
current pointer points to the node with which we are currently working.
Questions 20 195

3. Allocate memory for the root node:

root = (struct town *) malloc (sizeof (struct town));
Be sure to check for errors. The variable root will be a null pointer if no memory could
be allocated for the node.
4. Initialize the members of the root node:
root->north = NULL;
root->south = NULL;
root->east = NULL;
root->west = NULL;
strcpy (root->name, "New Haven");
Note that NULL pointers tell the program when it has come to the edge of the data
structure, that is, when it has found a link that doesn’t lead anywhere. At the moment,
the links of the root node do not point anywhere. This will change as we add more
nodes to the data structure.
5. Create a new, non-root node:
current = (struct town *) malloc (sizeof (struct town));
6. Initialize the current node:
current->north = NULL;
current->south = root;
current->east = NULL;
current->west = NULL;
strcpy (current->name, "North Haven");
7. Link neighboring nodes to the current node, as appropriate:
root->north = current;
8. Repeat steps 5 through 7, as necessary.
See Section 21.3 [Controlled recursion with data structures], page 200, for a practical
example of building a simple linked list programmatically.

20.5 Further data structure examples

See Chapter 24 [Example programs], page 215, to see a complete application program that
uses a complex data structure with nodes and links to stores its data later in the book.

20.6 Questions 20
1. What is the difference between a structure and a union?
2. What is a member?
3. If foo is a structure variable, how would you find out the value of its member bar?
4. If foo is a pointer to a structure variable, how would you find out the value of its
member bar?
5. How are data usually linked to make a complex data structure?
6. Every structure variable in a complex data structure must have its own variable name.
True or false?
7. How are the members of structures accessed in a data structure?
8. Write a small program to make linked list that contains three nodes long and set all
their values to be zero. Can you automate this program with a loop? Can you make it
work for any number of nodes?
196 Chapter 20: Data structures
The stack in detail 197

21 Recursion
The program that swallowed its tail.
This chapter is about functions that call themselves. Consider the program below:
#include <stdio.h>

void black_hole()
{
black_hole();
}

/* To shorten example, not using argp */

int main ()
{
black_hole();
return 0;
}
The main function calls the black_hole function, which calls itself, which calls itself,
which calls. . . Once the control flow enters black_hole, it will never exit. This kind of
function is called a recursive function, and a function’s act of calling itself is called recursion.

21.1 The stack

What happens when we run the last example program? The black_hole function calls
itself indefinitely. Each function call uses up a small portion of the computer’s memory
called the stack. Eventually all of this memory is used up, and a kind of error called a stack
overflow occurs. The program then crashes with a Segmentation fault error.
It is sometimes helpful to think of a function as a robot that does a job. A function
definition in effect provides the blueprints for a robot. When the function is executed, it
is as though a robot is built on an assembly line in a robot factory. A recursive function
is like a robot that builds a copy of itself on the same assembly line. The second robot is
identical to the first in every way, except that it is an assistant to the first robot, and has
been passed different arguments. This second robot may in turn build a copy of itself as
well, and so on. It is crucial that the process of robots building robots stop at some point;
otherwise, the robot factory will run out of raw materials (that is, computer memory), and
the assembly line will grind to a halt.

21.1.1 The stack in detail

Let’s examine this process in detail. When one function calls another function in a C
program, control passes from the first function to the second function. When the second
function ends, control passes back to the statement in the first function that immediately
follows the function call. But how does the computer know where in its memory this
statement resides?
The answer is simple. The computer keeps a list of the addresses in memory of the
places to which it must return, no matter how many function calls are made. This list is
the stack.
The stack gets its name from the fact that it is a LIFO, or last in, first out structure,
meaning that the last item to be pushed onto the stack is the first item to be popped off. It
198 Chapter 21: Recursion

works, in other words, like the stack of dinner plates you keep in your kitchen cabinet. As
you wash plates, you pile them one by one on top of the stack, and when you want a plate,
you take one from the top of the stack. The stack of plates in your cabinet is therefore also
a last in, first out structure, like the computer’s stack.
When one C function calls a second function, the computer leaves itself an address at
the top of the stack of where it should return when it has finished executing the second
function. If the second function calls a third function, the computer will push another
address onto the stack. When the third function has finished executing, the computer pops
the top address off the stack, which tells it where in the second function it should return.
When the second function has finished, the computer again pops the top address off the
stack — which tells it where in the first function it should return. Perhaps the first function
then calls another function, and the whole process starts again.
What happens when black_hole calls itself? The computer makes a note of the address
it must return to and pushes that address onto the top of the stack. It begins executing
black_hole again, and encounters another call to black_hole. The computer pushes
another address onto the top of the stack, and begins executing black_hole again. Since
the program has no chance of popping addresses off the stack, as the process continues, the
stack gets filled up with addresses. Eventually, the stack fills up and the program crashes.

21.2 Controlled recursion

If that were all there is to recursion, no one would ever use it. However, recursion can be
limited so it does not go out of control. Controlled recursion can be a powerful programming
technique.
When we discussed data structures, we remarked that programs and data structures
should aim to model the situation they deal with closely. Some structures, both in real life
and in computer memory, are made up of many levels of detail, and the details are roughly
the same at every level. For example, a genealogical tree starts with an individual with two
parents, each of whom has two parents, each of whom. . . These sorts of structure are called
self-similar.
Since recursion employs functions that contain calls to themselves, in effect creating
multiple self-similar levels of detail, controlled recursion is useful for dealing with self-similar
problems.
Recursive functions can be controlled by making sure that there is a safe way to exit
them at some point in the chain of function calls. The number of times recursion takes
place is limited by making a decision about whether the function calls itself or not. Simply
put, somewhere along the chain of function calls, the function makes the decision not to
call itself again, in a process nicknamed bottoming out. At that point, the program begins
popping addresses off the stack and returning to the previous functions. Eventually, the
very first function in the chain terminates, and the program ends successfully.
A standard example of controlled recursion is the factorial function. This is a mathe-
matical function which is important in statistics. The factorial function is defined to be
the product (multiplication) of all integers from 1 to the parameter of the function. (The
factorial of 0 is 1.)
Here are some examples of the factorial function. These are not executable C code
examples, but pseudocode:
Controlled recursion 199

factorial(3) == 1 * 2 * 3 == 6
factorial(4) == 1 * 2 * 3 * 4 == 24
factorial(3) == 1 * 2 * 3 * 4 * 5 == 120
Formally, the factorial function is defined by two equations. (Again, these are in pseu-
docode).
factorial(n) = n * factorial(n-1)
factorial(0) = 1
The first of these statements is recursive, because it defines the value of factorial(n)
in terms of factorial(n-1). The second statement allows the function to “bottom out”.
Here is a short code example that incorporates a factorial function.
#include <stdio.h>

int factorial (int n)

{
if (n == 0)
return 1;
else
return (n * factorial (n-1));
}

/* To shorten example, not using argp */

int main ()
{
printf ("%d\n", factorial(3));
return 0;
}
Let’s follow the control flow in this program to see how controlled recursion can work.
The main function prints the value of factorial(3). First, the factorial function is
called with the parameter 3. The function tests whether its parameter n is zero. It is not,
so it takes the else branch if the if statement, which instructs it to return the value of
factorial(3-1). It therefore calls itself recursively with a parameter of 2.
The new call checks whether its parameter is zero. It isn’t (it’s 2), so it takes the else
branch again, and tries to calculate 2 * factorial (1). In order to do so, it calls itself
recursively with a value of 2-1, or 1. The new call checks whether its parameter is zero. It
is actually 1, so it takes the else branch again and attempts to calculate 1 * factorial
(0). In order to do so, it calls itself again with the parameter 0.
Again, the function checks whether its parameter is zero. This time it is, so the function
bottoms out. It takes the first branch of the if statement and returns a value of 1. Now
the previous function call can also return a value, and so on, until the very first call to
factorial terminates, and the function returns a value of 6.
To sum up, the expression factorial(3) goes through the following steps before finally
being evaluated:
factorial (3) == 3 * factorial(2)
== 3 * (2 * factorial(1))
== 3 * (2 * (1 * factorial(0)))
== 3 * (2 * (1 * 1)))
== 6

Note: Make sure that the test for whether to bottom out your recursive function does
not depend on a global variable.
200 Chapter 21: Recursion

Suppose you have a global variable called countdown, which your recursive function
decrements by 1 every time it is called. When countdown equals zero, your recursive
function bottoms out. However, since other functions than the recursive function have
access to global variables, it is possible that another function might independently change
countdown in such a way that your recursive function would never bottom out — perhaps
by continually incrementing it, or perhaps even by setting it to a negative number.

21.3 Controlled recursion with data structures

Self-similar data structures are sometimes called recursive data structures. The simplest
recursive data structure is the linked list. At every node in a linked list, there is data of a
certain type and a link to the next node. The next simplest recursive data structure is the
binary tree, which splits into two branches at every node. Recursive functions be useful for
manipulating such recursive data structures.

The following code example makes use of recursion to print the value contained in the
last node in a linked list.

#include <stdio.h>

struct list_node
{
int data;
struct list_node *next;
};

struct list_node *last_node (struct list_node *node)

{
if (node->next == NULL)
return node;
else
return last_node (node->next);
}

/* To shorten example, not using argp */

int main ()
{
struct list_node *root;
struct list_node *current;
struct list_node *old;
struct list_node *last;

/* Initialize list. */
root = (struct list_node *) malloc (sizeof (struct list_node));
root->data = 1;
old = root;

current = (struct list_node *) malloc (sizeof (struct list_node));

current->data = 2;
old->next = current;
old = current;
Questions 21 201

current = (struct list_node *) malloc (sizeof (struct list_node));

current->data = 3;
old->next = current;
current->next = NULL;

/* Print data in last node. */

last = last_node (root);
printf ("Data in last node is %d.\n", last->data);

return 0;
}
This example program prints out the following line:
Data in last node is 3.
The last_node function, when passed a pointer to a node (such as the root), follows the
linked list to its end from that point, and returns a pointer to that node. It does so through
recursion. When it is passed a pointer to a node, it checks whether that node’s next link is
a null pointer. If the pointer is null, last_node has found the last node, and bottoms out,
returning a pointer to the current node; otherwise, it calls itself with a pointer to the next
node as a parameter.

21.4 Recursion summary

Recursion can be a powerful programming technique, especially when dealing with math-
ematical functions such as factorialisation that lend themselves naturally to recursion, or
with self-similar data structures. There is a major disadvantage to recursion, however, and
that is the amount of memory required to make it work. Do not forget that the program
stack grows each time a function call is made. If a function calls itself too many times, your
program will run out of memory and crash. Recursive programming can also be difficult;
runaway recursion is a common error. Therefore, be judicious in your use of recursion.

21.5 Questions 21
1. What is a recursive function?
2. What is a program stack, and what is it for?
3. State the major disadvantage of recursion.
202 Chapter 21: Recursion
Comments and style 203

22 Style

C has no rules about when to start new lines, where to place whitespace, and so on. Users
are free to choose a style which best suits them, but unless a strict style is adopted, sloppy
programs tend to result.
In older compilers, memory restrictions sometimes necessitated bizarre, cryptic styles
in the interest of efficiency. However, contemporary compilers such as GCC have no such
restrictions, and have optimizers that can produce faster code than most programmers
could write themselves by hand, so there are no excuses not to write programs as clearly as
possible.
No simple set of rules will ever provide a complete methodology for writing good pro-
grams. In the end, experience and good judgment are the factors which determine whether
you will write good programs. Nevertheless, a few guidelines to good style can be stated.
Many of the guidelines in this chapter are the distilled wisdom of countless C program-
mers over the decades that C has existed, and some come directly from the GNU Coding
Standards. That document contains more good advice than can be crammed into this short
chapter, so if you plan to write programs for the Free Software Foundation, you are urged
to consult section “Table of Contents” in GNU Coding Standards.

22.1 Formatting code

Place the open curly bracket that starts the body of a C function in the first column of
your source file, and avoid placing any other open brackets or open parentheses in that
column. This will help many code-processing utilities find the beginnings of your functions.
Similarly, you should also place the name of your functions within your function definitions
in the first column. Thus, your functions should resemble the following example:
static char *
concat (char *s1, char *s2)
{
...
}

When you split an expression into multiple lines, split it before an operator, not after
one. Here is the right way:
if (foo_this_is_long && bar > win (x, y, z)
&& remaining_condition)

Don’t declare multiple variables in one declaration that spans lines. Start a new decla-
ration on each line instead. For example, instead of this:
int foo,
bar;

write either this:

int foo, bar;

or this:
int foo;
int bar;
204 Chapter 22: Style

22.2 Comments and style

Comments are crucial for other programmers trying to understand your code. Every pro-
gram should start with a comment saying briefly what it is for. Example: ‘fmt - filter
for simple filling of text’. Similarly, you should put a comment on each function say-
ing what the function does, what sort of arguments it takes, what the possible values of
arguments mean, and what they are used for.
Please write all comments in a GNU program in English, because English is the one
language that nearly all programmers in all countries can read.

22.3 Variable and function names

The names of variables and functions in a program serve as comments of a sort, so try
to give your variables descriptive names (for example, num_of_books, cost_per_entry, or
distance_from_center). Names should be in English, like other comments.
Use underscores rather than internal capitalization in names, so that Emacs word
commands can be useful within them — thus distance_from_center rather than
distanceFromCenter or DistanceFromCenter. In fact, upper-case letters should be
reserved for macros and enum constants. Macros should be completely in upper case, for
example STANDARD_SIZE.
It used to be common practice to use the same local variables (with names like temp)
over and over for different purposes within one function. Instead, it is better to declare a
separate local variable for each distinct purpose, and give it a meaningful name. This not
only makes programs easier to understand, it also facilitates optimization by good compilers.

22.4 Declarations and initialization

You should explicitly declare the types of all objects. For example, explicitly declare all
arguments to functions, and declare all function that return integers to return type int,
even though the ANSI Standard permits omitting the int.
If there are only a few declarations, then initializing variables where you declare them
can be tidy, but if there are many variables to declare, then it is usually better to declare and
initialize separately, for the sake of clarity. In a long function, it is often good to initialize
the variable near where you are using it, so that someone reading the code does not have
to hunt around in the function to discover its initial value. (See Section 5.3 [Initialization],
page 20.)

22.5 Global variables and style

Global variables have caused almost as much controversy as the goto statement. Some
programmers say you should never use them. Other programmers use them on a regular
basis. In fact, while global variables should not be overused, they can simplify your code
considerably. The following guidelines may help you decide where to use globals.
• Always think of using local variables first. Global variables can puncture the encap-
sulization of your functions, that is, the logical isolation of your functions from the
rest of your code. It is difficult to see what variables are being passed to a function
unless they are all passed as parameters, so it is easier to debug a program when
encapsulization is maintained.
Questions 22 205

• Local variables may be impractical, however, if they mean passing the same dozen
parameters to multiple functions; in such cases, global variables will often streamline
your code.
• Data structures that are important to the whole program should be defined globally.
In “real programs” such as GNU Emacs, there are far more global variables than there
are local variables visible in any one function.
Finally, don’t use local variables or parameters that have the same names as global
identifiers. This can make debugging very difficult.

22.6 Hidden operators and style

Hiding operators away inside other statements can certainly make programs look elegant
and compact, but it can make programs harder to understand. Never forget that besides
being a set of instructions to the computer, programming is a form of communication to
other programmers. Be kind to the reader of your program. It could be you in months or
years to come.
Statements such as:
if ((my_int = (int)my_char++) <= --my_int2)
{
...
}
are not good style, and are no more efficient than the more longwinded:
my_int = (int) my_char;
my_char++;
my_int2--;

if (my_int <= my_int2)

{
...
}

22.7 Final words on style

It is easy to support pre-ANSI-Standard compilers in most programs, so if you know how
to do that and a program you are maintaining has such support, you should try to keep it
working.
Whatever style you use, use it consistently. A mixture of styles within a single program
tends to look ugly and be hard to read and maintain. If you are contributing changes to an
existing program, it is best to follow the style of that program.

22.8 Questions 22
1. Where should the name of a program and the opening bracket of a function definition
begin?
2. In what human language should comments be written for the GNU Project? Why?
3. Which is better as the name of a variable: plotArea, PlotArea, or plot_area? Why?
4. Why is it important to initialize a variable near where it is used in a long function?
5. Give an example of a case where using local variables is impractical.
206 Chapter 22: Style
parse error at. . . , parse error before. . . 207

23 Debugging

True artificial intelligence has not yet been achieved. C compilers are not intelligent, but
unconscious: mechanical in the derogatory sense of the word. Therefore, debugging your
programs can be a difficult process. A single typographical error can cause a compiler to
completely misunderstand your code and generate a misleading error message. Sometimes
a long string of compiler error messages are generated because of a single error in your code.
To minimize the time you spend debugging, it is useful to become familiar with the most
common compiler messages and their probable causes.
The first section in this chapter lists some of these common compile-time errors and
what to do about them. The next two sections discuss run-time errors in general, and
mathematical errors in particular. The final section introduces GDB, the GNU Debugger,
and explains some simple steps you can take to debug your programs with it.

23.1 Compile-time errors

In this section, we will examine a variety of compile-time errors and what you can do about
them. The aim is not to be a comprehensive guide to everything that can go wrong with
your program and all the corresponding error messages, but rather to give you a taste of
the kinds of errors you are likely to make, and to build your confidence by showing that
even fairly scary-looking error messages often have a simple cause.

23.1.1 parse error at. . . , parse error before. . .

This is a general-purpose syntax error. It is frequently caused by a missing semicolon. For
example, the following code:
#include <stdio.h>

/* To shorten example, not using argp */

int main()
{
printf ("Hello, world!\n")
return 0;
}

generates the following error:

semicolon.c: In function ‘main’:
semicolon.c:6: parse error before ‘return’

Adding a semicolon (‘;’) at the end of the line printf ("Hello, world!") will get rid
of this error.
Notice that the error refers to line 6, but the error is actually on the previous line. This
is quite common. Since C compilers are lenient about where you place whitespace, the
compiler treats line 5 and line 6 as a single line that reads as follows:
printf ("Hello, world!\n") return 0;

Of course this code makes no sense, and that is why the compiler complains.
Often a missing curly bracket will cause one of these errors. For example, the following
code:
208 Chapter 23: Debugging

#include <stdio.h>

/* To shorten example, not using argp */

int main()
{
if (1==1)
{
printf ("Hello, world!\n");

return 0;
}

generates the following error:

brackets.c: In function ‘main’:
brackets.c:11: parse error at end of input

Because there is no closing curly bracket for the if statement, the compiler thinks the
curly bracket that terminates the main function actually terminates the if statement. When
it does not find a curly bracket on line 11 of the program to terminate the main function,
it complains. One way to avoid this problem is to type both members of a matching pair
of brackets before you fill them in.

23.1.2 undefined reference to. . .

This error is often generated because you have typed the name of a function or variable
incorrectly. For example, the following code:
#include <stdio.h>

void print_hello()
{
printf ("Hello!\n");
}

/* To shorten example, not using argp */

int main()
{
Print_hello();
return 0;
}

generates the following rather forbidding error:

/tmp/cc9KXhmV.o: In function ‘main’:
/tmp/cc9KXhmV.o(.text+0x1f): undefined reference to ‘Print_hello’
collect2: ld returned 1 exit status

The answer, however, is very simple. C is case-sensitive. The main function calls the
function Print_hello (with a capital ‘P’), but the correct name of the function is print_
hello (with a lower-case ‘p’). The linker could not find a function with the name Print_
hello.

23.1.3 unterminated string or character constant

This error is often generated by code like the following:
different type arg 209

#include <stdio.h>

/* To shorten example, not using argp */

int main()
{
printf("hello!\n);
printf("Hello again!\n");
return 0;
}

The actual error message received was:

missquotes.c:6: unterminated string or character constant
missquotes.c:5: possible real start of unterminated constant

The compiler never found a close quote (‘"’) for the string ‘Hello!\n’. It read all
the text up from the quote in the line printf("Hello!\n); to the first quote in the line
printf("Hello again!\n"); as a single string. Notice that GCC helpfully suggests that it
is line 5 that actually contains the unterminated string. GCC is pretty smart as C compilers
go.

23.2 . . . undeclared (first use in this function)

This is similar to the ‘undefined reference to...’ error, but instead of referring to an
undefined function, you are referring to an undefined variable.
Sometimes this is a scope problem. You might get this error if you tried to refer to
another function’s local variable. For example:
#include <stdio.h>

void set_value()
{
int my_int = 5;
}

/* To shorten example, not using argp */

int main()
{
my_int = 23;
return 0;
}

The variable my_int is local to the function set_value, so referring to it from within main
results in the following error:
undec.c: In function ‘main’:
undec.c:10: ‘my_int’ undeclared (first use in this function)
undec.c:10: (Each undeclared identifier is reported only once
undec.c:10: for each function it appears in.)

23.2.1 different type arg

You might get this warning if you mismatch a parameter to printf and a conversion
specifier. For example, the following code:
210 Chapter 23: Debugging

#include <stdio.h>

/* To shorten example, not using argp */

int main()
{
int my_int = 5;
printf ("%f", my_int);
return 0;
}
produces the folliwing warning:
wrongtype2.c: In function ‘main’:
wrongtype2.c:6: warning: double format, different type arg (arg 2)
The ‘%f’ conversion specifier requires a floating-point argument, while my_int is an
integer, so GCC complains.
Note: GCC is quite lenient about type mismatches and will usually coerce one type
to another dynamically without complaining, for example when assigning a floating-point
number to an integer. This extends to mismatched parameters and conversion specifiers —
although you may receive odd results from printf and so on, the causes of which may not
be obvious. Therefore, in order to generate this warning, the ‘-Wall’ option of GCC was
used. This option causes GCC to be especially sensitive to errors, and to complain about
problems it usually ignores. You will often find the ‘-Wall’ option to be useful in finding
tricky problems. Here is the actual command line used to compile this program:
gcc -Wall -o wrong wrongtype2.c

23.2.2 too few parameters. . . , too many parameters. . .

Consider the following program:
#include <stdio.h>

void tweedledee (int a, int b, int c)

{
}

void tweedledum (int a, int b)

{
}

/* To shorten example, not using argp */

int main()
{

tweedledee (1, 2);

tweedledum (1, 2, 3);

return 0;
}
The tweedledee function takes three parameters, but main passes it two, whereas the
tweedledum function takes two parameters, but main passes it three. The result is a pair
of straightforward error messages:
params.c: In function ‘main’:
params.c:14: too few arguments to function ‘tweedledee’
params.c:15: too many arguments to function ‘tweedledum’
Unwarranted assumptions about storage 211

This is one reason for the existence of function prototypes. Before the ANSI Standard,
compilers did not complain about this kind of error. If you were working with a library
of functions with which you were not familiar, and you passed one the wrong number of
parameters, the error was sometimes difficult to track. Contemporary C compilers such as
GCC that follow the standard make finding parameter mismatch errors simple.

23.3 Run-time errors

This section examines errors that cannot be caught by the compiler and are exhibited only
when the program is run.

23.3.1 Confusion of = and ==

Consider the following program:
#include <stdio.h>

/* To shorten example, not using argp */

int main()
{
int my_int = 0;

if (my_int = 1)
{
printf ("Hello!\n");
}

return 0;
}
What will this program do? If you guessed that it will print ‘Hello!’, you are correct.
The assignment operator (=) was used by mistake instead of the equality operator (==).
What is being tested in the above if statement is not whether my_int has a value of 1
(which would be written if my_int == 1), but instead what the value is of the assignment
statement my_int = 1. Since the value of an assignment statement is always the result of
the assignment, and my_int is here being assigned a value of 1, the result is 1, which C
considers to be equivalent to TRUE. Thus, the program prints out its greeting.
Even the best C programmers make this mistake from time to time, and tracking down
an error like this can be maddening. Using the ‘-Wall’ option of GCC can help at least a
little by giving you a warning like the following:
equals.c: In function ‘main’:
equals.c:7: warning: suggest parentheses around assignment used as truth value

23.3.2 Confusing foo++ and ++foo

In many cases, the forms foo++ and ++foo are identical. However, if they are hidden inside
another statement, there can be a subtle difference. For example:
my_array [++my_index] = 0;
The code ++my_index cause my_index to be incremented by 1 before the assignment
takes place, whereas my_index++ would have cause my_index to be incremented after the
assignment takes place. Sometimes you’ll want one and sometimes the other. If you find
that your program is miscalculating by a difference of 1 (this is called an off-by-one bug
and is quite common), a prefix or postfix ‘++’ could be the cause. The same holds for other
prefix and postfix operators, such as ‘--’.
212 Chapter 23: Debugging

23.3.3 Unwarranted assumptions about storage

Do not assume that the size of a structure is the sum of the sizes of its parts. The two may
differ for various reasons; for example, the operating system may be aligning variables with
specific addresses within the data structure. Furthermore, the elements of an array may
not even be next to one another in memory.
This kind of code is always safe:
int my_array[3];

my_array[0] = 0;
my_array[1] = 0;
my_array[2] = 0;

This kind of code is not:

int my_array[3];

*my_array = 0;
*(my_array + (1 * sizeof(int))) = 0;
*(my_array + (2 * sizeof(int))) = 0;

While it is true that the variable my_array used without its square brackets is a pointer
to the first element of the array, you must not assume that you can simply calculate a
pointer to the third element with code like the following:
my_array + 2 * sizeof(int);

Do something like this instead:

&(my_array[2]);

23.3.4 Array out of bounds

When you get or set the value of an element of an array, GCC does not check whether you
are working within the bound of the array. In the worst case, this can lead to your program
crashing (but probably nothing worse happening on a GNU system). See Section 14.1
[Array bounds], page 86, for more information on this error. See Section 23.5 [Introduction
to GDB], page 213, for information on how you can check whether you are violating array
bounds, using the GNU Debugger.

23.3.5 Uncoordinated output

You may occasionally experience strange effects when writing output to the screen, such as
no output at all until the input is complete, or spontaneous bursts of output at seemingly
random intervals. This sort of problem usually has to do with the way the output is buffered.
The solution is usually to write a newline character (‘\n’) to the output when you are ready
to display it, or to use a function like fflush to flush the buffer. (See Section 16.1.5 [Stream
buffering], page 111, for more information.)

23.3.6 Global variables and recursion

Global variables and recursion usually do not mix. Make sure that the test for whether to
“bottom out” your recursive function does not depend on a global variable. See Section 21.2
[Controlled recursion], page 198, for more information on why this is a bad thing.
Questions 23 213

23.4 Mathematical errors

Mathematical errors are a special kind of run-time error. They may not necessarily cause
your program to crash, but they are likely to produce all sorts of strange results if you are
doing some complex calculations in your program. Consider the following line of code:
root = sqrt (-1.0);
Readers with a smattering of mathematics will recognise that this code cannot give a
sensible answer. The square root of -1 is a complex number called i. The number i is a
so-called imaginary number, and cannot be represented by a floating-point value, which is
what the sqrt function returns.
What happens in such a case? Two things:
1. The value returned is a special floating-point macro such as NAN (which means “not a
number”) or INFINITY.
2. More importantly from a debugging standpoint, a floating-point exception occurs. An
exception is an error condition, and when a floating-point exception is raised, as the jar-
gon goes, an error flag is set in the operating system, signifying what kind of exception
it was (in other words, what kind of error caused the exception to be raised).
There are several kinds of floating-point exception:
• FE_INVALID: The “Invalid Operation” exception. Raised if the operands are invalid for
the given operation, for example, if you are trying to take the square root of a negative
number, as above.
• FE_DIVBYZERO: The “Division by Zero” exception. Raised when a finite, nonzero
number is divided by zero.
• FE_OVERFLOW: The “Overflow” exception. Raised when the result cannot be expressed
as a finite value, for example when a finite, nonzero number is divided by zero. When-
ever this exception is raised, the FE_INEXACT exception is also raised.
• FE_UNDERFLOW: The “Underflow” exception. Raised when an intermediate result is too
small to be calculated accurately, or when an operation’s rounded result is too small to
be normalized. Normalisation, roughly speaking, is the process of converting a number
to scientific notation, such as converting 235 to 2.35e2, where the mantissa, or number
to the left of the ‘e’, must not be zero. See Section 5.1.2 [Floating point variables],
page 18, for more information on scientific notation.)
• FE_INEXACT: The “Inexact” exception. Raised if a rounded result is not exact, for
example when calculating an irrational number such as the square root of 2.
You can test for these exceptions with the fetestexcept function, which takes one
parameter, a bitwise OR’d list of the exception flags from the list above for which you
are testing, and returns a nonzero value containing a bitwise OR’d list of the flags you
passed it for the exceptions that actually occurred. You can also clear selected flags with
the feclearexcept function, which accepts a bitwise-OR’d list of exception flags to clear,
and returns zero if it was successful. (You can pass either of these function the macro
FE_ALL_EXCEPT, which contains all of the floating-point exception flags OR’d together.)
In case this explanation is unclear, let’s look at a practical example.

23.5 Introduction to GDB

Blah blah blah.
214 Chapter 23: Debugging

23.6 Questions 23
Spot the errors in the following:
Blah blah blah.
Example programs 215

24 Example programs
The aim of this section is to provide a substantial example of C programming, using in-
put from and output to disk, GNU-style long options, and the linked list data structure
(including insertion, deletion, and sorting of nodes).
#include <stdio.h>
#include <string.h>
#include <argp.h>

#define NAME_LEN 100

#define ADDR_LEN 500

const char *argp_program_version =

"bigex 1.0";

const char *argp_program_bug_address =

"<bug-gnu-utilsgnu.org>";

/* This structure is used by main to communicate with parse_opt. */

struct arguments
{
char *args[1]; /* No arguments to this function */
int verbose; /* The -v flag */
char *infile; /* Argument for -i */
char *outfile; /* Argument for -o */
};

struct personal_data
{
char name[NAME_LEN];
char address[ADDR_LEN];
struct personal_data *next;
};

/*
OPTIONS. Field 1 in ARGP.
Order of fields: {NAME, KEY, ARG, FLAGS, DOC}.
*/
static struct argp_option options[] =
{
{"verbose", ’v’, 0, 0, "Produce verbose output"},

{"input", ’i’, "INFILE", 0,

"Read addresses from INFILE"},

{"output", ’o’, "OUTFILE", 0,

"Output to OUTFILE instead of to standard output"},

{0}
};

/*
PARSER. Field 2 in ARGP.
216 Chapter 24: Example programs

Order of parameters: KEY, ARG, STATE.

*/
static error_t
parse_opt (int key, char *arg, struct argp_state *state)
{
struct arguments *arguments = state->input;

switch (key)
{
case ’v’:
arguments->verbose = 1;
break;
case ’i’:
arguments->infile = arg;
break;
case ’o’:
arguments->outfile = arg;
break;
case ARGP_KEY_ARG:
if (state->arg_num >= 1)
{
argp_usage(state);
}
arguments->args[state->arg_num] = arg;
break;
case ARGP_KEY_END:
if (state->arg_num < 1)
{
argp_usage (state);
}
break;
default:
return ARGP_ERR_UNKNOWN;
}
return 0;
}

/*
ARGS_DOC. Field 3 in ARGP.
A description of the non-option command-line arguments
that we accept.
*/
static char args_doc[] = "ARG";

/*
DOC. Field 4 in ARGP.
Program documentation.
*/
static char doc[] =
"bigex -- Add ARG new names to an address book file.\vThe largest code example in the GNU C Tutorial.";

/*
The ARGP structure itself.
*/
static struct argp argp = {options, parse_opt, args_doc, doc};
Example programs 217

struct personal_data *
new_empty_node()
{
struct personal_data *new_node;

new_node = (struct personal_data*)

malloc (sizeof (struct personal_data));

strcpy (new_node->name, "");

strcpy (new_node->address, "");
new_node->next = NULL;

return new_node;
}

struct personal_data *
create_node()
{
int bytes_read;
int nbytes;

struct personal_data *current_node;

char *name;
char *address;

current_node = new_empty_node();

puts ("Name?");
nbytes = NAME_LEN;
name = (char *) malloc (nbytes + 1);
bytes_read = getline (&name, &nbytes, stdin);
if (bytes_read == -1)
{
puts ("ERROR!");
}
else
{
strncpy (current_node->name, name, NAME_LEN);
free (name);
}

puts ("Address?");
nbytes = ADDR_LEN;
address = (char *) malloc (nbytes + 1);
bytes_read = getline (&address, &nbytes, stdin);
if (bytes_read == -1)
{
puts ("ERROR!");
}
else
{
strncpy (current_node->address, address, ADDR_LEN);
free (address);
}
218 Chapter 24: Example programs

printf("\n");
return current_node;
}

struct personal_data *
find_end_node (struct personal_data *current_node)
{
if (current_node->next == NULL)
{
return current_node;
}
else
{
return find_end_node (current_node->next);
}
}

int
list_length (struct personal_data *root)
{
struct personal_data *current_node;
int count = 0;

current_node = root;

while (current_node->next != NULL)

{
current_node = current_node->next;
count++;
}
return count;
}

struct personal_data *
find_node (struct personal_data *root,
int node_wanted)
{
struct personal_data *current_node;
int index = 0;

current_node = root;

while ((index < node_wanted) && (current_node->next != NULL))

{
current_node = current_node->next;
index++;
}
return current_node;
}

delete_node (struct personal_data *root,

int location)
Example programs 219

{
struct personal_data *previous_node;
struct personal_data *current_node;

previous_node = find_node (root, location - 1);

current_node = find_node (root, location);
previous_node->next = current_node->next;
}

insert_node (struct personal_data *root,

struct personal_data *new_node,
int location)
{
struct personal_data *temp_ptr;
struct personal_data *previous_node;

previous_node = find_node (root, location - 1);

temp_ptr = previous_node->next;

previous_node->next = new_node;
new_node->next = temp_ptr;
}

swap_nodes (struct personal_data *root, int a, int b)

{
int temp;
struct personal_data *node_a;
struct personal_data *node_b;
struct personal_data *temp_node;

if (a > b)
{
temp = a;
a = b;
b = temp;
}

node_b = find_node (root, b);

delete_node (root, b);

node_a = find_node (root, a);

delete_node (root, a);

insert_node (root, node_b, a);

insert_node (root, node_a, b);
}

sort_list (struct personal_data *root)

{
int i, j, list_len, diff;

list_len = list_length (root);

for (i=2; i<=list_len; i++)
{
220 Chapter 24: Example programs

j = i;
while (strcmp ( (find_node(root, j))->name,
(find_node(root, j-1))->name) < 0)
{
swap_nodes (root, j, j-1);
j--;
}
}
}

print_node (struct personal_data *current_node,

FILE *save_stream)
{
fprintf (save_stream, "%s%s",
current_node->name,
current_node->address);
}

print_list (struct personal_data *current_node,

FILE *save_stream)
{
print_node (current_node, save_stream);

if (current_node->next != NULL)
{
print_list (current_node->next, save_stream);
}
}

struct personal_data *
read_node (FILE *instream)
{
int bytes_read;
int nbytes;

struct personal_data *current_node;

char *name;
char *address;
char *blankline;
int read_err = 0;

current_node = new_empty_node();

nbytes = NAME_LEN;
name = (char *) malloc (nbytes + 1);
bytes_read = getline (&name, &nbytes, instream);
if (bytes_read == -1)
{
read_err = 1;
}
else
{
puts (name);
Example programs 221

strncpy (current_node->name, name, NAME_LEN);

free (name);
}

nbytes = ADDR_LEN;
address = (char *) malloc (nbytes + 1);
bytes_read = getline (&address, &nbytes, instream);
if (bytes_read == -1)
{
read_err = 1;
}
else
{
puts (address);
strncpy (current_node->address, address, ADDR_LEN);
free (address);
}

if (read_err)
{
return NULL;
}
else
{
return current_node;
}
}

struct personal_data *
read_file (char *infile)
{
FILE *input_stream = NULL;
struct personal_data *root;
struct personal_data *end_node;
struct personal_data *current_node;

root = new_empty_node();
end_node = root;

input_stream = fopen (infile, "r");

if (input_stream)
{
while (current_node = read_node (input_stream))
{
end_node->next = current_node;
end_node = current_node;
end_node->next = NULL;
}
}
return root;
}

/*
The main function.
Notice how now the only function call needed to process
222 Chapter 24: Example programs

all command-line options and arguments nicely

is argp_parse.
*/
int
main (int argc, char **argv)
{
struct arguments arguments;
struct personal_data *root;
struct personal_data *end_node;
struct personal_data *current_node;
int i, newnum;
FILE *save_stream;

/* Set argument defaults */

arguments.infile = NULL;
arguments.outfile = NULL;
arguments.verbose = 0;

/* Where the magic happens */

argp_parse (&argp, argc, argv, 0, 0, &arguments);

if (arguments.infile)
{
root = read_file (arguments.infile);
end_node = find_end_node (root);
}
else
{
root = new_empty_node();
end_node = root;
}

/* Where do we send output? */

if (arguments.outfile)
save_stream = fopen (arguments.outfile, "w");
else
save_stream = stdout;

newnum = atoi (arguments.args[0]);

for (i = 1; i <= newnum; i++)

{
current_node = create_node();
end_node->next = current_node;
end_node = current_node;
end_node->next = NULL;
}

sort_list (root);
print_list (root->next, save_stream);

/* Close stream; skip error-checking for brevity of example */

fclose (save_stream);

/* If in verbose mode, print song stanza */

if (arguments.verbose)
{};
Example programs 223

return 0;
}
224 Chapter 24: Example programs
A note from the original author 225

Appendix A A note from the original author

This book began life in 1987 as one of the early books on C programming. I wrote it during
a summer vacation from University, in England in 1987. It was published by Dabs Press, a
small publishing house which specialized in books for microcomputers, particularly Acorn’s
classic BBC micro. With the arrival of the Amiga, I was able to obtain a C compiler. I had
had my eye on C for some time, and I felt at the time, that it was the best language I had
seen to date for system programming. The publisher and I decided that C would become
the ‘BASIC’ of the 16-bit microcomputer world, which it did. C took off, and the book sold
very well for a number of years. As the contract expired, the book was lost in my files, until
I met Richard Stallman and he asked me if I would give the tutorial to GNU. I agreed to
recover the original files from diskettes and partly re-work them, to remove the emphasis
from micro-computers and over to GNU. The result of that work was the first version of
the tutorial. Having handed over the merchandise, I agreed that it would be a good thing
for others to update and improve the tutorial. My only requirement was that I would be
allowed to explain a few changes for which I would not want to be blamed. I cannot insist
that such changes will not be made, but I can at least distance myself from them. They
are fairly picky and even silly things, but for one reason or another, they mean a lot to me.
The first has to do with grammar. The grammar which is written and printed in books
today is often incorrect. Many colloquialisms and vernacular perversions of grammar are
printed and believed to be correct. I am fairly sure that no such errors are mine! The other
thing has to do with the style and formatting of C code. The placement of curly braces is
something about which I agree with only a handful of people on the planet. Kernighan and
Ritchie’s original placement of curly braces is so horrendous that I would go so far as to
call it "wrong", logically and aesthetically. The GNU indentation, which positions braces
of the same level in straight vertical alignment, is much better, but in my opinion it gets its
indentation wrong. I would indent 3 positions before the first brace, and keep the text within
braces aligned with the braces themselves, rather than indenting within the braces, as GNU
does. That, in my personal opinion, makes it easier to identify text and braces belonging
together, and leads to optimal clarity. I also insist that curly braces be used around single
statements, in loops and tests, even when not strictly required by the language. Finally,
having grown up in England and lived in Norway, which have contradictory punctuation
rules, I am utterly confused about punctuation and have probably made many errors. With
that little spiel said, I now pass the torch to future authors and wish everyone luck. I am
happy to see an old summer job not go to waste.
Mark Burgess, Oslo March 2001
226 Appendix A: A note from the original author
Reserved words in C 227

Appendix B Reserved words in C

Blah blah blah.
Here is a list of all the reserved words in C. The set of reserved words above is used to
build up the basic instructions of C; you can not use them in programs your write
Please note that this list is somewhat misleading. Many more words are out of bounds.
This is because most of the facilities which C offers are in libraries that are included in
programs. Once a library has been included in a program, its functions are defined and you
cannot use their names yourself.
C requires all of these reserved words to be in lower case. (This does mean that, typed
in upper case, the reserved words could be used as variable names, but this is not recom-
mended.)
(A "d" by the word implies that it is used as part of a declaration.)
auto d if
break int d
case long d
char d register d
continue return
default short d
do sizeof
double d static d
else struct
entry switch
extern d typedef d
float d union d
for unsigned d
goto while

also in modern implementations:

enum d
void d

const d
signed d
volatile d
228 Appendix B: Reserved words in C
Precedence of operators 229

Appendix C Precedence of operators

The highest priority operators are listed first.
Operator Operation Evaluated

() parentheses left to right

[] square brackets left to right

++ increment right to left

-- decrement right to left
(type) cast operator right to left
* the contents of right to left
& the address of right to left
- unary minus right to left
~ one’s complement right to left
! logical NOT right to left

* multiply left to right

/ divide left to right
% remainder (MOD) left to right

+ add left to right

- subtract left to right

>> shift right left to right

<< shift left left to right

> is greater than left to right

>= greater than or equal to left to right
<= less than or equal to left to right
< less than left to right

== is equal to left to right

!= is not equal to left to right

& bitwise AND left to right

^ bitwise exclusive OR left to right
| bitwsie includive OR left to right
&& logical AND left to right
|| logical OR left to right

= assign right to left

+= add assign right to left
-= subtract assign right to left
*= multiply assign right to left
/= divide assign right to left
%= remainder assign right to left
>>= right shift assign right to left
<<= left shift assign right to left
&= AND assign right to left
^= exclusive OR assign right to left
|= inclusive OR assign right to left
230 Appendix C: Precedence of operators
Special characters 231

Appendix D Special characters

Control characters are invisible on the screen. They have special purposes usually to do
with cursor movement and are written into an ordinary string or character by typing a
backslash character \ followed by some other character. These characters are listed below.
A character can be any ASCII character, printable or not printable from values -128 to
127. (But only 0 to 127 are used.) Control characters i.e. non printable characters are put
into programs by using a backslash \ and a special character or number. The characters
and their meanings are:
‘\b’ backspace BS
‘\f’ form feed FF (also clear screen)
‘\n’ new line NL (like pressing return)
‘\r’ carriage return CR (cursor to start of line)
‘\t’ horizontal tab HT
‘\v’ vertical tab (not all versions)
‘\x’ ???
‘\"’ double quotes (not all versions)
‘\’’ single quote character ’
‘\\’ backslash character \
‘\ddd ’ character ddd where ddd is an ASCII code given in octal or base 8. (See
Appendix C)
Here is a code example that prints special characters:
/***************************************************/
/* */
/* Special Characters */
/* */
/***************************************************/

#include <stdio.h>

main ()

{
printf ("Beep! \7 \n");
printf ("ch = \’a\’ \n");
printf (" <- Start of this line!! \r");
}
The output of this program is:
Beep! (and the BELL sound)
ch = ’a’
<- Start of this line!!
and the text cursor is left where the arrow points.
232 Appendix D: Special characters
Character conversion table 233

Appendix E Character conversion table

This table lists the decimal, octal, and hexadecimal numbers for characters 0 – 127.
Decimal Octal Hexadecimal Character

0 0 0 CTRL-@
1 1 1 CTRL-A
2 2 2 CTRL-B
3 3 3 CTRL-C
4 4 4 CTRL-D
5 5 5 CTRL-E
234 Appendix E: Character conversion table

6 6 6 CTRL-F
7 7 7 CTRL-G
8 10 8 CTRL-H
9 11 9 CTRL-I
10 12 A CTRL-J
11 13 B CTRL-K
12 14 C CTRL-L
13 15 D CTRL-M
14 16 E CTRL-N
15 17 F CTRL-O
16 20 10 CTRL-P
17 21 11 CTRL-Q
18 22 12 CTRL-R
19 23 13 CTRL-S
20 24 14 CTRL-T
21 25 15 CTRL-U
22 26 16 CTRL-V
23 27 17 CTRL-W
24 30 18 CTRL-X
25 31 19 CTRL-Y
26 32 1A CTRL-Z
27 33 1B CTRL-[
28 34 1C CTRL-\
29 35 1D CTRL-]
30 36 1E CTRL-^
31 37 1F CTRL-_
32 40 20
33 41 21 !
34 42 22 "
35 43 23 #
36 44 24 $
37 45 25 %
38 46 26 &
39 47 27 ’
40 50 28 (
41 51 29 )
42 52 2A *
43 53 2B +
44 54 2C ,
45 55 2D -
46 56 2E .
47 57 2F /
48 60 30 0
49 61 31 1
50 62 32 2
51 63 33 3
52 64 34 4
53 65 35 5
54 66 36 6
55 67 37 7
56 70 38 8
57 71 39 9
58 72 3A :
59 73 3B ;
60 74 3C <
61 75 3D =
62 76 3E >
63 77 3F ?
64 100 40 @
65 101 41 A
66 102 42 B
67 103 43 C
68 104 44 D
69 105 45 E
70 106 46 F
71 107 47 G
A word about goto 235

Appendix F A word about goto

This word is redundant in C and encourages poor programming style. For this reason it
has been ignored in this book. For completeness, and for those who insist on using it (may
their programs recover gracefully) the form of the goto statement is as follows:
goto label;
label is an identifier which occurs somewhere else in the given function and is defined as a
label by using the colon:
label : printf ("Ugh! You used a goto!");
236 Appendix F: A word about goto
Answers to questions 237

Appendix G Answers to questions

Blah blah blah.
238 Appendix G: Answers to questions
Bibliography 239

Bibliography
Blah blah blah.
240 Bibliography
Glossary 241

Glossary
Blah blah blah.
242 Glossary
Code index 243

Code index

(Index is nonexistent)
244 Code index
Concept index 245

Concept index

(Index is nonexistent)
246 Concept index
Assigning variables to one another 247

Bits and pieces

This section is for random chunks of text that are too good to drop from the book, but
were out-of-place where they were.

Allocating memory for strings

Neither of the methods above is any good if a program is going to be fetching a lot of strings
from a user. It just isn’t practical to define lots of static strings and expect the user to type
into the right size boxes! The next step in string handling is therefore to allocate memory
for strings personally: in other words to be able to say how much storage is needed for a
string while a program is running. C has special memory allocation functions which can do
this, not only for strings but for any kind of object. Suppose then that a program is going
to get ten strings from the user. Here is one way in which it could be done:
1. Define one large, static string (or array) for getting one string at a time. Call this a
string buffer, or waiting place.
2. Define an array of ten pointers to characters, so that the strings can be recalled easily.
3. Find out how long the string in the string buffer is.
4. Allocate memory for the string.
5. Copy the string from the buffer to the new storage and place a pointer to it in the
array of pointers for reference.
6. Release the memory when it is finished with.

Characters
In C, single characters are written enclosed by single quotes. This is in contrast to strings
of characters, which use double quotes (‘"..."’).
int ch;
ch = ’a’;
would give ch the value of the character ‘a’. The same effect can also be achieved by writing:
char ch = ’a’;
It is also possible to have the type:
unsigned char
This admits ASCII values from 0 to 255, rather than -128 to 127.

Assigning variables to one another

Not only can you assign numbers to variables, you can assign other variables to variables:
var1 = 23;
var2 = var1;
The variable or value on either side of the ‘=’ symbol must usually be of the same type.
However, integers and characters will interconvert because characters are stored by their
ASCII codes (which are integers!) Thus the following will work:
int i;
char ch = ’A’;

i = ch;
248 Bits and pieces

printf ("The ASCII code of %c is %d",ch,i);

The result of this would be:
The ASCII code of A is 65

Function pointers
You can create pointers to functions as well as to variables. Function pointers can be tricky,
however, and caution is advised in using them.
Function pointers allow you to pass functions as a parameters to another function. This
enables you to give the latter function a choice of functions to call. That is, you can plug
in a new function in place of an old one simply by passing a different parameter. This
technique is sometimes called indirection or vectoring.
To pass a pointer for one function to a second function, simply use the name of the
first function, as long as there is no variable with the same name. Do not include the first
function’s parentheses or parameters when you pass its name.
For example, the following code passes a pointer for the function named fred_function
to the function barbara_function:
void fred();
barbara (fred);
Notice that fred is declared with a regular function prototype before barbara calls it. You
must also declare barbara, of course:
void barbara (void (*function_ptr)() );
Notice the parentheses around function_ptr and the parentheses after it. As far as barbara
is concerned, any function passed to it is named (*function_ptr)(), and this is how fred
is called in the example below:
#include <stdio.h>

void fred();
void barbara ( void (*function_ptr)() );
int main();

int main()
{
barbara (fred);
return 0;
}

void fred()
{
printf("fred here!\n");
}

void barbara ( void (*function_ptr)() )

{
/* Call fred */
(*function_ptr)();
}
The output from this example is simply ‘fred here!’.
Again, notice how barbara called fred. Given a pointer to a function, the syntax for
calling the function is as follows:
Function pointers 249

variable = (*function_pointer )(parameter_list );

For example, in the program below, the function do_math calls the functions add and
subtract with the following line:
result = (*math_fn_ptr) (num1, num2);

Here is the example program:

#include <stdio.h>

int add (int, int);

int subtract (int, int);
int do_math (int (*math_fn_ptr) (int, int), int, int);
int main();

int main()
{
int result;

result = do_math (add, 10, 5);

printf ("Addition = %d.\n", result);

result = do_math (subtract, 40, 5);

printf ("Subtraction = %d.\n\n", result);

return 0;
}

int add (int num1, int num2)

{
return (num1 + num2);
}

int subtract (int num1, int num2)

{
return (num1 - num2);
}

int do_math (int (*math_fn_ptr) (int, int), int num1, int num2)
{
int result;

printf ("\ndo_math here.\n");

/* Call one of the math functions passed to us:

either add or subtract. */

result = (*math_fn_ptr) (num1, num2);

return result;
}

The output from this program reads:

do_math here.
Addition = 15.

do_math here.
Subtraction = 35.
250 Bits and pieces

You can also initialize a function pointer by setting it to the name of a function, then
treating the function pointer as an ordinary function, as in the next example:
#include <stdio.h>

int main();
void print_it();
void (*fn_ptr)();

int main()
{
void (*fn_ptr)() = print_it;

(*fn_ptr)();

return 0;
}

void print_it()
{
printf("We are here! We are here!\n\n");
}
Remember to initialize any function pointers you use this way! If you do not, your
program will probably crash, because the uninitialized function pointer will contain garbage.
Table of Contents i

Table of Contents

Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xi

1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 The advantages of C . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Questions for Chapter 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

2 Using a compiler . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.1 Basic ideas about C . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.2 The compiler. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.3 File names . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.4 Errors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.4.1 Typographical errors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.4.2 Type errors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.5 Questions for Chapter 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

3 The form of a C program. . . . . . . . . . . . . . . . . . . 7

3.1 A word about style . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
3.2 Comments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
3.3 Example 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
3.4 Questions for Chapter 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

4 Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
4.1 Function names . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
4.2 Function examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
4.3 Functions with values . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
4.4 Function prototyping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
4.5 The exit function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
4.6 Questions for Chapter 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

5 Variables and declarations . . . . . . . . . . . . . . . . . 17

5.1 Integer variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
5.1.1 The char type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
5.1.2 Floating point variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
5.2 Declarations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
5.3 Initialization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
5.4 The cast operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
5.4.1 Cast operator demo . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
5.5 Storage classes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
5.5.1 External variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
5.5.2 Static variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
5.5.3 Other storage classes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
5.6 Questions for Chapter 5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
ii

6 Scope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
6.1 Global Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
6.2 Local Variables. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
6.3 Communication via parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
6.4 Scope example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
6.5 Questions for Chapter 6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

7 Expressions and operators . . . . . . . . . . . . . . . . . 29

7.1 The assignment operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
7.1.1 Important note about assignment . . . . . . . . . . . . . . . . . . . . . . . 30
7.2 Expressions and values . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
7.3 Expressions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
7.4 Parentheses and Priority . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
7.5 Unary Operator Precedence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
7.6 Special Assignment Operators ++ and -- . . . . . . . . . . . . . . . . . . . . . 32
7.7 More Special Assignments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
7.8 Comparisons and logic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
7.9 Logical operators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
7.9.1 Inclusive OR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
7.10 Questions for Chapter 7 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

8 Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
8.1 Parameters in function prototypes . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
8.2 Value Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
8.3 Actual parameters and formal parameters . . . . . . . . . . . . . . . . . . . . 39
8.4 Variadic functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
8.5 Questions for Chapter 8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

9 Pointers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
9.1 Pointer operators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
9.2 Pointer types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
9.3 Pointers and initialization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
9.4 Variable parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
9.4.1 Passing pointers correctly . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
9.4.2 Another variable parameter example . . . . . . . . . . . . . . . . . . . . 48
9.5 Questions for Chapter 9 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

10 Decisions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
10.1 if . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
10.2 if... else... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
10.3 Nested if statements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
10.4 The ?. . . :. . . operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
10.5 The switch statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
10.6 Example Listing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
10.7 Questions for Chapter 10 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
Table of Contents iii

11 Loops . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
11.1 while . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
11.2 do. . . while . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
11.3 for . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
11.4 The flexibility of for . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
11.5 Terminating and speeding loops . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
11.5.1 Terminating loops with break . . . . . . . . . . . . . . . . . . . . . . . . . 64
11.5.2 Terminating loops with return . . . . . . . . . . . . . . . . . . . . . . . . 64
11.5.3 Speeding loops with continue . . . . . . . . . . . . . . . . . . . . . . . . . 65
11.6 Nested loops . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
11.7 Questions for Chapter 11 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66

12 Preprocessor directives. . . . . . . . . . . . . . . . . . . 67
12.1 A few directives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
12.2 Macros . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
12.2.1 Macro functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
12.3 Extended macro example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
12.4 Questions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72

13 Libraries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
13.1 Header files . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
13.2 Kinds of library . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
13.3 Common library functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
13.3.1 Character handling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
13.4 Mathematical functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
13.5 Questions for Chapter 13 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83

14 Arrays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
14.1 Array bounds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
14.2 Arrays and for loops . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
14.3 Multidimensional arrays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
14.4 Arrays and nested loops . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
14.5 Initializing arrays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
14.6 Arrays as Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
14.7 Questions for Chapter 14 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95

15 Strings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
15.1 Conventions and declarations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
15.2 Initializing strings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
15.3 String arrays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
15.4 String library functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
15.5 Questions for Chapter 15 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
iv

16 Input and output . . . . . . . . . . . . . . . . . . . . . . . 103

16.1 High-level file routines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
16.1.1 Opening a file . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
16.1.2 Closing a file . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
16.1.3 Block input and output . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
16.1.4 File position. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
16.1.5 Stream buffering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
16.1.6 End-of-file and error functions . . . . . . . . . . . . . . . . . . . . . . . . 112
16.2 String output and input . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
16.2.1 Unformatted string output . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
16.2.1.1 puts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
16.2.1.2 fputs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
16.2.2 Formatted string output . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
16.2.2.1 printf . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
16.2.2.2 Formatted output conversion specifiers . . . . . . . . . . . . 115
16.2.3 fprintf . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
16.2.4 asprintf . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
16.2.5 Deprecated formatted string output functions . . . . . . . . . . 119
16.2.5.1 sprintf . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
16.2.6 String input . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
16.2.6.1 getline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
16.2.6.2 getdelim . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
16.2.7 Deprecated string input functions . . . . . . . . . . . . . . . . . . . . . 121
16.2.7.1 gets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
16.2.7.2 fgets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
16.2.8 Formatted string input . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
16.2.8.1 sscanf . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
16.2.8.2 Formatted input conversion specifiers . . . . . . . . . . . . . 125
16.2.9 Deprecated formatted string input functions . . . . . . . . . . . 128
16.2.9.1 scanf . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
16.2.9.2 String overflows with scanf . . . . . . . . . . . . . . . . . . . . . . . 128
16.2.10 fscanf . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
16.3 Single-character input and output . . . . . . . . . . . . . . . . . . . . . . . . . 130
16.3.1 getchar . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
16.3.2 putchar . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
16.3.3 getc and fgetc . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
16.3.4 putc and fputc . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
16.3.5 ungetc() . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
16.4 Programming with pipes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
16.5 Low-level file routines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
16.5.1 Usual file name errors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
16.5.2 Opening files at a low level . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
16.5.2.1 File creation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
16.5.3 Closing files at a low level . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
16.5.4 Reading files at a low level . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
16.5.5 Writing files at a low level . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
16.5.6 Finding file positions at a low level . . . . . . . . . . . . . . . . . . . . 143
16.5.7 Deleting files at a low level . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
Table of Contents v

16.5.8 Renaming files at a low level. . . . . . . . . . . . . . . . . . . . . . . . . . 144

16.6 Questions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145

17 Putting a program together . . . . . . . . . . . . . 147

17.1 argc and argv . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
17.2 Processing command-line options . . . . . . . . . . . . . . . . . . . . . . . . . . 148
17.2.1 argp description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
17.2.2 argp example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
17.3 Environment variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
17.4 Compiling multiple files. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
17.5 Writing a makefile. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
17.5.1 What a Rule Looks Like . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158
17.5.2 A simple makefile . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
17.5.3 make in action . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160
17.5.4 Variables simplify makefiles. . . . . . . . . . . . . . . . . . . . . . . . . . . 161
17.5.5 Letting make deduce commands . . . . . . . . . . . . . . . . . . . . . . 161
17.5.6 Combining rules by prerequisite. . . . . . . . . . . . . . . . . . . . . . . 162
17.5.7 Rules for cleaning the directory . . . . . . . . . . . . . . . . . . . . . . . 163
17.6 Building a library . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
17.7 Questions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165

18 Advanced operators . . . . . . . . . . . . . . . . . . . . 167

18.1 Hidden operators and values . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
18.1.1 Hidden assignments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
18.1.2 Postfix and prefix ++ and -- . . . . . . . . . . . . . . . . . . . . . . . . . . 168
18.1.3 Arrays and hidden operators . . . . . . . . . . . . . . . . . . . . . . . . . 169
18.1.4 A warning about style . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
18.2 The comma operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
18.3 Machine-level operators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
18.3.1 Bitwise operators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
18.3.2 Shift operations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
18.3.3 Truth tables and bit masks . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
18.3.3.1 Bitwise NOT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
18.3.3.2 Bitwise AND . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
18.3.3.3 Bitwise inclusive OR . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
18.3.3.4 Bitwise exclusive OR (XOR/EOR) . . . . . . . . . . . . . . . 173
18.3.3.5 Masks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
18.4 Questions 18 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
vi

19 More data types . . . . . . . . . . . . . . . . . . . . . . . . 177

19.1 enum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177
19.2 void . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179
19.3 volatile . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
19.4 Constants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
19.4.1 const. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
19.4.2 Constant expressions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
19.5 struct and union. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
19.6 typedef . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
19.7 Questions 19 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182

20 Data structures. . . . . . . . . . . . . . . . . . . . . . . . . 183

20.1 struct . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
20.1.1 Structure declarations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
20.1.1.1 Structure declarations using typedef . . . . . . . . . . . . . . 184
20.1.2 Using structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184
20.1.3 Arrays of structures. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
20.1.4 Nested structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186
20.1.5 Pointers to structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
20.1.6 Initializing structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
20.2 Memory allocation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189
20.3 union . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190
20.3.1 Declaration of unions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190
20.3.2 Using unions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
20.4 Complex data structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192
20.4.1 Data structure diagrams . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192
20.4.2 Dynamic data structures, Pointers and Dynamic Memory
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192
20.4.3 Lists and trees . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193
20.4.3.1 Linked lists. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193
20.4.3.2 Binary trees . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194
20.4.4 Setting up a data structure . . . . . . . . . . . . . . . . . . . . . . . . . . . 194
20.4.4.1 Designing your data structure . . . . . . . . . . . . . . . . . . . . 194
20.4.4.2 Initializing your data structure . . . . . . . . . . . . . . . . . . . 194
20.5 Further data structure examples . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
20.6 Questions 20 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195

21 Recursion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
21.1 The stack. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
21.1.1 The stack in detail . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
21.2 Controlled recursion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198
21.3 Controlled recursion with data structures . . . . . . . . . . . . . . . . . . 200
21.4 Recursion summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
21.5 Questions 21 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
Table of Contents vii

22 Style . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203
22.1 Formatting code . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203
22.2 Comments and style . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204
22.3 Variable and function names . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204
22.4 Declarations and initialization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204
22.5 Global variables and style . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204
22.6 Hidden operators and style. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
22.7 Final words on style . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
22.8 Questions 22 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205

23 Debugging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
23.1 Compile-time errors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
23.1.1 parse error at. . . , parse error before. . . . . . . . . . . . . . . . . . 207
23.1.2 undefined reference to. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 208
23.1.3 unterminated string or character constant . . . . . . . . . . . . . 208
23.2 . . . undeclared (first use in this function) . . . . . . . . . . . . . . . . . . . 209
23.2.1 different type arg . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209
23.2.2 too few parameters. . . , too many parameters. . . . . . . . . . 210
23.3 Run-time errors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
23.3.1 Confusion of = and ==. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
23.3.2 Confusing foo++ and ++foo . . . . . . . . . . . . . . . . . . . . . . . . . . 211
23.3.3 Unwarranted assumptions about storage . . . . . . . . . . . . . . . 212
23.3.4 Array out of bounds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212
23.3.5 Uncoordinated output . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212
23.3.6 Global variables and recursion . . . . . . . . . . . . . . . . . . . . . . . . 212
23.4 Mathematical errors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
23.5 Introduction to GDB . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
23.6 Questions 23 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214

24 Example programs . . . . . . . . . . . . . . . . . . . . . . 215

Appendix A A note from the original author

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225

Appendix B Reserved words in C . . . . . . . . . . 227

Appendix C Precedence of operators . . . . . . . 229

Appendix D Special characters. . . . . . . . . . . . . 231

Appendix E Character conversion table . . . . 233

Appendix F A word about goto . . . . . . . . . . . . 235

viii

Appendix G Answers to questions. . . . . . . . . . 237

Bibliography. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 239

Glossary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 241

Code index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243

Concept index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245

Bits and pieces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247

Allocating memory for strings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247
Characters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247
Assigning variables to one another . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247
Function pointers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 248