C Pocket Reference – Read Now and Download Mobi

C Pocket Reference

Peter Prinz

Ulla Kirch-Prinz

Editor

Jonathan Gennick

Copyright © 2009 O'Reilly Media, Inc.

O'Reilly Media

Chapter 1. C Pocket Reference

Introduction

The programming language C was developed in the 1970s by Dennis Ritchie at Bell Labs (Murray Hill, New Jersey) in the process of implementing the Unix operating system on a DEC PDP-11 computer. C has its origins in the typeless programming language BCPL (Basic Combined Programming Language, developed by M. Richards) and in B (developed by K. Thompson). In 1978, Brian Kernighan and Dennis Ritchie produced the first publicly available description of C, now known as the K&R standard.

C is a highly portable language oriented towards the architecture of today's computers. The actual language itself is relatively small and contains few hardware-specific elements. It includes no input/output statements or memory management techniques, for example. Functions to address these tasks are available in the extensive C standard library.

C's design has significant advantages:

• Source code is highly portable

• Machine code is efficient

• C compilers are available for all current systems

The first part of this pocket reference describes the C language, and the second part is devoted to the C standard library. The description of C is based on the ANSI X3.159 standard. This standard corresponds to the international standard ISO/IEC 9899, which was adopted by the International Organization for Standardization in 1990, then amended in 1995 and 1999. The ISO/IEC 9899 standard can be ordered from the ANSI web site; see http://ansi.org/public/std_info.html.

The 1995 standard is supported by all common C compilers today. The new extensions defined in the 1999 release (called "ANSI C99" for short) are not yet implemented in many C compilers, and are therefore specially labeled in this book. New types, functions, and macros introduced in ANSI C99 are indicated by an asterisk in parentheses ^(*).

Font Conventions

The following typographic conventions are used in this book:

Italic

Used to introduce new terms, and to indicate filenames.

Constant width

Used for C program code as well as for functions and directives.

Constant width italic

Indicates replaceable items within code syntax.

Constant width bold

Used to highlight code passages for special attention.

Fundamentals

A C program consists of individual building blocks called functions , which can invoke one another. Each function performs a certain task. Ready-made functions are available in the standard library; other functions are written by the programmer as necessary. A special function name is main( ): this designates the first function invoked when a program starts. All other functions are subroutines.

C Program Structure

Figure 1-1 illustrates the structure of a C program. The program shown consists of the functions main() and showPage() , and prints the beginning of a text file to be specified on the command line when the program is started.

Figure 1-1. A C program

The statements that make up the functions, together with the necessary declarations and preprocessing directives, form the source code of a C program. For small programs, the source code is written in a single source file . Larger C programs consist of several source files, which can be edited and compiled separately. Each such source file contains functions that belong to a logical unit, such as functions for output to a terminal, for example. Information that is needed in several source files, such as declarations, is placed in header files. These can then be included in each source file via the #include directive.

Source files have names ending in .c; header files have names ending in .h. A source file together with the header files included in it is called a translation unit .

There is no prescribed order in which functions must be defined. The function showPage() in Figure 1-1 could also be placed before the function main(). A function cannot be defined within another function, however.

The compiler processes each source file in sequence and decomposes its contents into tokens , such as function names and operators. Tokens can be separated by one or more whitespace characters, such as space, tab, or newline characters. Thus only the order of tokens in the file matters. The layout of the source code—line breaks and indentation, for example—is unimportant. The preprocessing directives are an exception to this rule, however. These directives are commands to be executed by the preprocessor before the actual program is compiled, and each one occupies a line to itself, beginning with a hash mark (#).

Comments are any strings enclosed either between /* and */, or between // and the end of the line. In the preliminary phases of translation, before any object code is generated, each comment is replaced by one space. Then the preprocessing directives are executed.

Character Sets

ANSI C defines two character sets. The first is the source character set , which is the set of characters that may be used in a source file. The second is the execution character set , which consists of all the characters that are interpreted during the execution of the program, such as the characters in a string constant.

Each of these character sets contains a basic character set , which includes the following:

• The 52 upper- and lower-case letters of the Latin alphabet:

  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

• The ten decimal digits (where the value of each character after 0 is one greater than the previous digit):

  0  1  2  3  4  5  6  7  8  9

• The following 29 graphic characters:

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

• The five whitespace characters:

  space, horizontal tab, vertical tab, newline, form feed

In addition, the basic execution character set contains the following:

• The null character \0, which terminates a character string

• The control characters represented by simple escape sequences , shown in Table 1-1, for controlling output devices such as terminals or printers

Table 1-1. The standard escape sequences

Any other characters, depending on the given compiler, can be used in comments, strings, and character constants. These may include the dollar sign or diacriticals, for example. However, the use of such characters may affect portability.

The set of all usable characters is called the extended character set , which is always a superset of the basic character set.

Certain languages use characters that require more than one byte. These multibyte characters may be included in the extended character set. Furthermore, ANSI C99 provides the integer type wchar_t (wide character type), which is large enough to represent any character in the extended character set. The modern Unicode character encoding is often used, which extends the standard ASCII code to represent some 35,000 characters from 24 countries.

C99 also introduces trigraph sequences . These sequences, shown in Table 1-2, can be used to input graphic characters that are not available on all keyboards. The sequence ??!, for example, can be entered to represent the "pipe" character |.

Table 1-2. The trigraph sequences

Identifiers

Identifiers are names of variables, functions, macros, types, etc. Identifiers are subject to the following formative rules:

• An identifier consists of a sequence of letters (A to Z, a to z), digits (0 to 9), and underscores (_).

• The first character of an identifier must not be a digit.

• Identifiers are case-sensitive.

• There is no restriction on the length of an identifier. However, only the first 31 characters are generally significant.

Keywords are reserved and must not be used as identifiers. Following is a list of keywords:

External names —that is, identifiers of externally linked functions and variables—may be subject to other restrictions, depending on the linker: in portable C programs, external names should be chosen so that only the first eight characters are significant, even if the linker is not case-sensitive.

Some examples of identifiers are:

Categories and Scope of Identifiers

Each identifier belongs to exactly one of the following four categories:

• Label names

• The tags of structures, unions, and enumerations. These are identifiers that follow one of the keywords struct, union, or enum (see Section 1.10).

• Names of structure or union members. Each structure or union type has a separate name space for its members.

• All other identifiers, called ordinary identifiers.

Identifiers of different categories may be identical. For example, a label name may also be used as a function name. Such re-use occurs most often with structures: the same string can be used to identify a structure type, one of its members, and a variable; for example:

struct person {char *person; /*...*/} person;

The same names can also be used for members of different structures.

Each identifier in the source code has a scope . The scope is that portion of the program in which the identifier can be used. The four possible scopes are:

Function prototype

Identifiers in the list of parameter declarations of a function prototype (not a function definition) have function prototype scope . Because these identifiers have no meaning outside the prototype itself, they are little more than comments.

Function

Only label names have function scope . Their use is limited to the function block in which the label is defined. Label names must also be unique within the function. The goto statement causes a jump to a labelled statement within the same function.

Block

Identifiers declared in a block that are not labels have block scope . The parameters in a function definition also have block scope. Block scope begins with the declaration of the identifier and ends with the closing brace (}) of the block.

File

Identifiers declared outside all blocks and parameter lists have file scope . File scope begins with the declaration of the identifier and extends to the end of the source file.

An identifier that is not a label name is not necessarily visible throughout its scope. If an identifier with the same category as an existing identifier is declared in a nested block, for example, the outer declaration is temporarily hidden. The outer declaration becomes visible again when the scope of the inner declaration ends.

Basic Types

The type of a variable determines how much space it occupies in storage and how the bit pattern stored is interpreted. Similarly, the type of a function determines how its return value is to be interpreted.

Types can be either predefined or derived. The predefined types in C are the basic types and the type void. The basic types consist of the integer types and the floating types.

Integer Types

There are five signed integer types: signed char , short int (or short), int , long int (or long), and long long int ^(*) (or long long ^(*)). For each of these types there is a corresponding unsigned integer type with the same storage size. The unsigned type is designated by the prefix unsigned in the type specifier, as in unsigned int.

The types char , signed char, and unsigned char are formally different. Depending on the compiler settings, however, char is equivalent either to signed char or to unsigned char. The prefix signed has no meaning for the types short, int, long, and long long ^(*), however, since they are always considered to be signed. Thus short and signed short specify the same type.

The storage size of the integer types is not defined; however, their width is ranked in the following order: char <= short <= int <= long <= long long ^(*). Furthermore, the size of type short is at least 2 bytes, long at least 4 bytes, and long long at least 8 bytes. Their value ranges for a given implementation are found in the header file limits.h .

ANSI C99 also introduces the type _Bool to represent Boolean values. The Boolean value true is represented by 1 and false by 0. If the header file stdbool.h has been included, then bool can be used as a synonym for _Bool and the macros true and false for the integer constants 1 and 0. Table 1-3 shows the standard integer types together with some typical value ranges.

Table 1-3. Standard integer types with storage sizes and value ranges

ANSI C99 introduced the header file stdint.h(*) , which defines integer types with specific widths (see Table 1-4). The width N of an integer type is the number of bits used to represent values of that type, including the sign bit. (Generally, N = 8, 16, 32, or 64.)

Table 1-4. Integer types with defined width

For example, int16_t is an integer type that is exactly 16 bits wide, and int_fast32_t is the fastest integer type that is 32 or more bits wide. These types must be defined for the widths N = 8, 16, 32, and 64. Other widths, such as int24_t, are optional. For example:

int16_t val = -10;  // integer variable
                    // width: exactly 16 bits

For each of the signed types described above, there is also an unsigned type with the prefix u. uintmax_t , for example, represents the implementation's widest unsigned integer type.

Real and Complex Floating Types

Three types are defined to represent non-integer real numbers: float , double , and long double . These three types are called the real floating types .

The storage size and the internal representation of these types are not specified in the C standard, and may vary from one compiler to another. Most compilers follow the IEEE 754-1985 standard for binary floating-point arithmetic, however. Table 1-5 is also based on the IEEE representation.

Table 1-5. Real floating types

The header file float.h defines symbolic constants that describe all aspects of the given representation (see Section 1.17).

Internal representation of a real floating-point number

The representation of a floating-point number x is always composed of a sign s, a mantissa m, and an exponent exp to base 2:

x = s * m * 2^exp, where 1.0 <= m < 2  or  m = 0

The precision of a floating type is determined by the number of bits used to store the mantissa. The value range is determined by the number of bits used for the exponent.

Figure 1-2 shows the storage format for the float type (32-bit) in IEEE representation.

Figure 1-2. IEEE storage format for the 32-bit float type

The sign bit S has the value 1 for negative numbers and 0 for other numbers. Because in binary the first bit of the mantissa is always 1, it is not represented. The exponent is stored with a bias added, which is 127 for the float type.

For example, the number -2.5 = -1 * 1.25 * 2¹ is stored as:

S = 1, Exponent = 1+127 = 128, Mantissa = 0.25

Complex floating types

ANSI C99 introduces special floating types to represent the complex numbers and the pure imaginary numbers. Every complex number z can be represented in Cartesian coordinates as follows:

z = x + i*y

where x and y are real numbers and i is the imaginary unit

.

The real numbers x and y represent respectively the real part and the imaginary part of z.

Complex numbers can also be represented in polar coordinates :

z = r * (cos(theta) + i * sin(theta))

The angle theta is called the argument and the number r is the magnitude or absolute value of z.

In C, a complex number is represented as a pair of real and imaginary parts, each of which has type float, double, or long double. The corresponding complex floating types arefloat _Complex, double _Complex, and long double _Complex.

In addition, the pure imaginary numbers—i. e., the complex numbers z = i*y where y is a real number—can also be represented by the types float _Imaginary , double _Imaginary , and long double _Imaginary .

Together, the real and the complex floating types make up the floating types.

The Type void

The type specifier void indicates that no value is available. It is used in three kinds of situations:

Expressions of type void

There are two uses for void expressions. First, functions that do not return a value are declared as void. For example:

void exit (int status);

Second, the cast construction (void)expression can be used to explicitly discard the value of an expression. For example:

(void)printf("An example.");

Prototypes of functions that have no parameters

For example:

int rand(void);

Pointers to void

The type void * (pronounced "pointer to void") represents the address of an object, but not the object's type. Such "typeless" pointers are mainly used in functions that can be called with pointers to different types as parameters. For example:

void *memcpy(void *dest, void *source, size_t count);

Constants

Every constant is either an integer constant, a floating constant, a character constant, or a string literal. There are also enumeration constants, which are described in Section 1.10.1. Every constant has a type that is determined by its value and its notation.

Integer Constants

Integer constants can be represented as ordinary decimal numbers, octal numbers, or hexadecimal numbers:

• A decimal constant (base 10) begins with a digit that is not 0; for example: 1024

• An octal constant (base 8) begins with a 0; for example: 012

• A hexadecimal constant (base 16) begins with the two characters 0x or 0X; for example: 0x7f, 0X7f, 0x7F, 0X7F. The hexadecimal digits A to F are not case-sensitive.

The type of an integer constant, if not explicitly specified, is the first type in the appropriate hierarchy that can represent its value.

For decimal constants, the hierarchy of types is:

int, long, unsigned long, long long(*).

For octal or hexadecimal constants, the hierarchy of types is:

int, unsigned int, long, unsigned long, long long(*), 
unsigned long long(*).

Thus, integer constants normally have type int. The type can also be explicitly specified by one of the suffixes L or l (for long), LL ^(*) or ll ^(*) (for long long ^(*)), and/or U or u (for unsigned). Table 1-6 provides some examples.

Table 1-6. Examples of integer constants

The macros in Table 1-7 are defined to represent constants of an integer type with a given maximum or minimum width N (e. g., = 8, 16, 32, 64). Each of these macros takes a constant integer as its argument and is replaced by the same value with the appropriate type.

Table 1-7. Macros for integer constants of minimum or maximum width

Floating Constants

A floating constant is represented as a sequence of decimal digits with one decimal point, or an exponent notation. Some examples are:

41.9 
5.67E-3   // The number  5.67*10^-3

E can also be written as e. The letter P or p is used to represent a floating constant with an exponent to base 2 (ANSI C99); for example:

2.7P+6    // The number  2.7*2⁶

The decimal point or the notation of an exponent using E, e, P ^(*), or p ^(*) is necessary to distinguish a floating constant from an integer constant.

Unless otherwise specified, a floating constant has type double. The suffix F or f assigns the constant the type float; the suffix L or l assigns it the type long double. Thus the constants in the previous examples have type double, 12.34F has type float, and 12.34L has type long double.

Each of the following constants has type double. All the constants in each row represent the same value:

Character Constants and String Literals

A character constant consists of one or more characters enclosed in single quotes. Some examples are:

'0'    'A'    'ab'

Character constants have type int. The value of a character constant that contains one character is the numerical value of the representation of the character. For example, in the ASCII code, the character constant '0' has the value 48, and the constant 'A' has the value 65.

The value of a character constant that contains more than one character is dependent on the given implementation. To ensure portability, character constants with more than one character should be avoided.

Escape sequences such as '\n' may be used in character constants. The characters ' and \ can also be represented this way.

The prefix L can be used to give a character constant the type wchar_t; for example:

L'A'    L'\x123'

A string literal consists of a sequence of characters and escape sequences enclosed in double quotation marks; for example:

"I am a string!\n"

A string literal is stored internally as an array of char (see Section 1.10) with the string terminator '\0'. It is therefore one byte longer than the specified character sequence. The empty string occupies exactly one byte. A string literal is also called a string constant, although the memory it occupies may be modified.

The string literal "Hello!", for example, is stored as a char array, as shown in Figure 1-3.

Figure 1-3. A string literal stored as a char array

String literals that are separated only by whitespace are concatenated into one string. For example:

Because the newline character is also a whitespace character, this concatenation provides a simple way to continue a long string literal in the next line of the source code.

Wide string literals can also be defined as arrays whose elements have type wchar_t. Again, this is done by using the prefix L; for example:

L"I am a string of wide characters!"

Expressions and Operators

An expression is a combination of operators and operands. In the simplest case, an expression consists simply of a constant, a variable, or a function call. Expressions can also serve as operands, and can be joined together by operators into more complex expressions.

Every expression has a type and, if the type is not void, a value. Some examples of expressions follow:

4 * 512                         // Type: int  
printf("An example!\n")         // Type: int 
1.0 + sin(x)                    // Type: double
srand((unsigned)time(NULL))     // Type: void
(int*)malloc(count*sizeof(int)) // Type: int *

In expressions with more than one operator, the precedence of the operators determines the grouping of operands with operators. The arithmetic operators *, /, and %, for example, take precedence over + and -. In other words, the usual rules apply for the order of operations in arithmetic expressions. For example:

4 + 6 * 512    // equivalent to 4 + (6 * 512)

If a different grouping is desired, parentheses must be used:

(4 + 6) * 512

Table 1-8 lists the precedence of operators.

Table 1-8. Precedence of operators

If two operators have equal precedence, then the operands are grouped as indicated in the "Grouping" column of Table 1-8. For example:

2 * 5 / 3     // equivalent to (2 * 5) / 3

Operators can be unary or binary : a unary operator has one operand, while a binary operator has two. This distinction is important for two reasons:

• All unary operators have the same precedence.

• The four characters -, +, *, and & can represent unary or binary operators, depending on the number of operands.

Furthermore, C has one ternary operator: the conditional operator ?: has three operands.

The individual operators are briefly described in Table 1-9 through Table 1-16 in the following sections. The order in which the operands are evaluated is not defined, except where indicated. For example, there's no guarantee which of the following functions will be invoked first:

f1() + f2() // Which of the two functions is 
            // called first is not defined.

Arithmetic Operators

Table 1-9. The arithmetic operators

The operands of arithmetic operators may have any arithmetic type. Only the % operator requires integer operands.

The usual arithmetic conversions may be performed on the operands. For example, 3.0/2 is equivalent to 3.0/2.0. The result has the type of the operands after such conversion.

Note that the result of division with integer operands is also an integer! For example:

6 / 4       //  Result:  1 
6 % 4       //  Result:  2 
6.0 / 4.0   //  Result: 1.5

The increment operator ++ (and analogously, the decrement operator --) can be placed either before or after its operand. A variable x is incremented (i. e., increased by 1) both by ++x (prefix notation ) and x++ (postfix notation ) . The expressions nonetheless yield different values: the expression ++x has the value of x increased by 1, while the expression x++ yields the prior, unincremented value of x.

Because the operators ++ and -- perform an assignment, their operand must be an lvalue; i. e., an expression that designates a location in memory, such as a variable.

The operators ++, --, + (addition), and - (subtraction) can also be used on pointers. For more information on pointers and pointer arithmetic, see Section 1.10.

Assignment Operators

Assignments are performed by simple and compound assignment operators, as shown in Table 1-10.

Table 1-10. Assignment operators

The left operand in an assignment must be an lvalue; i. e., an expression that designates an object. This object is assigned a new value.

The simplest examples of lvalues are variable names. In the case of a pointer variable ptr, both ptr and *ptr are lvalues. Constants and expressions such as x+1, on the other hand, are not lvalues.

The following operands are permissible in a simple assignment (=):

• Two operands with arithmetic types

• Two operands with the same structure or union type

• Two pointers that both point to objects of the same type, unless the right operand is the constant NULL

If one operand is a pointer to an object, then the other may be a pointer to the "incomplete" type void (i. e., void *).

If the two operands have different types, the value of the right operand is converted to the type of the left operand.

An assignment expression has the type and value of the left operand after the assignment. Assignments are grouped from right to left. For example:

a = b = 100;  // equivalent to  a=(b=100); 
              // The value 100 is assigned to b and a.

A compound assignment has the form x op= y, where op is a binary arithmetic operator or a binary bitwise operator. The value of x op (y) is assigned to x. For example:

a *= b+1;     // equivalent to  a = a * (b + 1);

In a compound assignment x op= y, the expression x is only evaluated once. This is the only difference between x op= y and x = x op (y).

Relational Operators and Logical Operators

Every comparison is an expression of type int that yields the value 1 or 0. The value 1 means "true" and 0 means "false." Comparisons use the relational operators listed in Table 1-11.

Table 1-11. The relational operators

The following operands are permissible for all relational operators:

• Two operands with real arithmetic types. The usual arithmetic conversions may be performed on the operands.

• Two pointers to objects of the same type.

The equality operators == and != can also be used to compare complex numbers. Furthermore, the operands may also be pointers to functions of the same type. A pointer may also be compared with NULL or with a pointer to void. For example:

int cmp, *p1, *p2;
. . .
cmp = p1 < p2; // if p1 is less than p2, then cmp = 1; 
               // otherwise cmp = 0.

Logical Operators

The logical operators, shown in Table 1-12, can be used to combine the results of several comparison expressions into one logical expression.

Table 1-12. The logical operators

The operands of logical operators may have any scalar (i. e., arithmetic or pointer) type. Any value except 0 is interpreted as "true"; 0 is "false."

Like relational expressions, logical expressions yield the values "true" or "false"; that is, the int values 0 or 1:

!x || y  // "(not x) or y" yields 1 (true)
         // if x == 0 or y != 0

The operators && and || first evaluate the left operand. If the result of the operation is already known from the value of the left operand (i. e., the left operand of && is 0 or the left operand of || is not 0), then the right operand is not evaluated. For example:

i < max  &&  scanf("%d", &x) == 1

In this logical expression, the function scanf() is only called if i is less than max.

Bitwise Operators

There are six bitwise operators, described in Table 1-13. All of them require integer operands.

Table 1-13. The bitwise operators

The logical bitwise operators, & (AND), | (OR), ^ (exclusive OR), and ~ (NOT) interpret their operands bit by bit: a bit that is set, i. e., 1, is considered "true"; a cleared bit, or 0, is "false". Thus, in the result of z = x & y, each bit is set if and only if the corresponding bit is set in both x and y. The usual arithmetic conversions are performed on the operands.

The shift operators << and >> transpose the bit pattern of the left operand by the number of bit positions indicated by the right operand. Integer promotions are performed beforehand on both operands. The result has the type of the left operand after integer promotion. Some examples are:

int  x = 0xF, result; 
result = x << 4; // yields 0xF0
result = x >> 2; // yields  0x3

The bit positions vacated at the right by the left shift << are always filled with 0 bits. Bit values shifted out to the left are lost.

The bit positions vacated at the left by the right shift >> are filled with 0 bits if the left operand is an unsigned type or has a non-negative value. If the left operand is signed and negative, the left bits may be filled with 0 (logical shift) or with the value of the sign bit (arithmetic shift), depending on the compiler.

Memory Accessing Operators

The operators in Table 1-14 are used to access objects in memory. The terms used here, such as pointer, array, structure, etc., are introduced later under Section 1.10.

Table 1-14. Memory accessing operators

The operand of the address operator & must be an expression that designates a function or an object. The address operator & yields the address of its operand. Thus an expression of the form &x is a pointer to x. The operand of & must not be a bit-field, nor a variable declared with the storage class specifier register.

The indirection operator * is used to access an object or a function through a pointer. If ptr is a pointer to an object or function, then *ptr designates the object or function pointed to by ptr. For example:

int  a, *pa;  // An int variable and a pointer to int. 

pa  = &a;     // Let pa point to a. 
*pa = 123;    // Now equivalent to  a = 123;

The subscript operator [] can be used to address the elements of an array. If v is an array and i is an integer, then v[i] denotes the element with index i in the array. In more general terms, one of the two operands of the operator [] must be a pointer to an object (e. g., an array name), and the other must be an integer. An expression of the form x[i] is equivalent to (*(x+(i))). For example:

float  a[10], *pa; // An array and a pointer.
pa = a;            // Let pa point to a[0].

Since pa points to a[0], pa[3] is equivalent to a[3] or *(a+3).

The operators . and -> designate a member of a structure or union. The left operand of the dot operator must have a structure or union type. The left operand of the arrow operator is a pointer to a structure or union. In both cases, the right operand is the name of a member of the type. The result has the type and value of the designated member.

If p is a pointer to a structure or union and x is the name of a member, then p->x is equivalent to (*p).x, and yields the member x of the structure (or union) to which p points.

The operators . and ->, like [], have the highest precedence, so that an expression such as ++p->x is equivalent to ++(p->x).

Other Operators

The operators in Table 1-15 do not belong to any of the categories described so far.

Table 1-15. Other operators

A function call consists of a pointer to a function (such as a function name) followed by parentheses () containing the argument list, which may be empty.

The cast operator can only be used on operands with scalar types! An expression of the form ( type )x yields the value of the operand x with the type specified in the parentheses.

The operand of the sizeof operator is either a type name in parentheses or any expression that does not have a function type. The sizeof operator yields the number of bytes required to store an object of the specified type, or the type of the expression. The result is a constant of type size_t.

The conditional operator ?: forms a conditional expression. In an expression of the form x?y:z, the left operand x is evaluated first. If the result is not equal to 0 (in other words, if x is "true"), then the second operand y is evaluated, and the expression yields the value of y. However, if x is equal to 0 ("false"), then the third operand z is evaluated, and the expression yields the value of z.

The first operand can have any scalar type. If the second and third operands do not have the same type, then a type conversion is performed. The type to which both can be converted is the type of the result. The following types are permissible for the second and third operands:

• Two operands with arithmetic types.

• Two operands with the same structure or union type, or the type void.

• Two pointers, both of which point to objects of the same type, unless one of them is the constant NULL. If one operand is an object pointer, the other may be a pointer to void.

The sequence or comma operator , has two operands: first the left operand is evaluated, then the right. The result has the type and value of the right operand. Note that a comma in a list of initializations or arguments is not an operator, but simply a punctuation mark!

Alternative notation for operators

The header file iso646.h defines symbolic constants that can be used as synonyms for certain operators, as listed in Table 1-16.

Table 1-16. Symbolic constants for operators

Type Conversions

A type conversion yields the value of an expression in a new type. Conversion can be performed only on scalar types, i. e., arithmetic types and pointers.

A type conversion always conserves the original value, if the new type is capable of representing it. Floating-point numbers may be rounded on conversion from double to float, for example.

Type conversions can be implicit —i. e., performed by the compiler automatically—or explicit , through the use of the cast operator. It is considered good programming style to use the cast operator whenever type conversions are necessary. This makes the type conversion explicit, and avoids compiler warnings.

Integer Promotion

Operands of the types _Bool, char, unsigned char, short, and unsigned short, as well as bit-fields, can be used in expressions wherever operands of type int or unsigned int are permissible. In such cases, integer promotion is performed on the operands: they are automatically converted to int or unsigned int. Such operands are converted to unsigned int only if the type int cannot represent all values of the original type.

Thus C always "expects" values that have at least type int. If c is a variable of type char, then its value in the expression:

 c + '0'

is promoted to int before the addition takes place.

Usual Arithmetic Conversions

The operands of a binary operator may have different arithmetic types. In this case, the usual arithmetic conversions are implicitly performed to cast their values in a common type. However, the usual arithmetic conversions are not performed for the assignment operators, nor for the logical operators && and ||.

If operands still have different types after integer promotion, they are converted to the type that appears highest in the hierarchy shown in Figure 1-4. The result of the operation also has this type.

Figure 1-4. Arithmetic type promotion hierarchy

When one complex floating type is converted to another, both the type of the real part and the type of the imaginary part are converted according to the rules applicable to the corresponding real floating types.

Type Conversions in Assignments and Pointers

A simple assignment may also involve different arithmetic types. In this case, the value of the right operand is always converted to the type of the left operand.

In a compound assignment, the usual arithmetic conversions are performed for the arithmetic operation. Then any further type conversion takes place as for a simple assignment.

A pointer to void can be converted to any other object pointer. An object pointer can also be converted into a pointer to void. The address it designates—its value—remains unchanged.

Statements

A statement specifies an action to be performed, such as an arithmetic operation or a function call. Many statements serve to control the flow of a program by defining loops and branches. Statements are processed one after another in sequence, except where such control statements result in jumps.

Every statement that is not a block is terminated by a semicolon.

Block and Expression Statements

A block , also called a compound statement, groups a number of statements together into one statement. A block can also contain declarations.

The syntax for a block is:

{[list of declarations][list of statements]}

Here is an example of a block:

{ int i = 0;      /* Declarations  */
  static long a;
  extern long max;

  ++a;            /* Statements    */
  if( a >= max)
  {   . . .    }  /* A nested block */
  . . .
}

The declarations in a block normally precede the statements. However, ANSI C99 permits free placement of declarations.

New blocks can occur anywhere within a function block. Usually a block is formed wherever the syntax calls for a statement, but the program requires several statements. This is the case, for example, when more than one statement is to be repeated in a loop.

An expression statement is an expression followed by a semicolon. The syntax is:

[expression] ;

Here is an example of an expression statement:

 y = x;        // Assignment

The expression—an assignment or function call, for example—is evaluated for its side effects. The type and value of the expression are discarded.

A statement consisting only of a semicolon is called an empty statement , and does not peform any operation. For example:

for ( i = 0;  str[i] != '\0'; ++i )
   ;                     // Empty statement

Jumps

The following statements can be used to control the program flow:

• Selection statements: if ... else or switch

• Loops: while, do ... while or for

• Unconditional jumps: goto, continue, break or return

Loops

A loop consists of a statement or block, called the loop body, that is executed several times, depending on a given condition. C offers three statements to construct loops: while, do ... while, and for.

In each of these loop statements, the number of loop iterations performed is determined by a controlling expression . This is an expression of a scalar type, i. e., an arithmetic expression or a pointer. The expression is interpreted as "true" if its value is not equal to 0; otherwise it is considered "false".

Syntactically, the loop body consists of one statement. If several statements are required, they are grouped in a block.

Unconditional Jumps

Name

if ... else

Synopsis

The if statement creates a conditional jump.

Syntax:

if (expression) 
                     statement1  [else  
                     statement2]

The expression must have a scalar type. First, the if statement's controlling expression is evaluated. If the result is not equal to 0—in other words, if the expression yields "true"—then statement1 is executed. Otherwise, if else is present, statement2 is executed.

Example:

if (x > y)  max = x; // Assign the greater of x and y to 
else        max = y; // the variable max.

The use of else is optional. If the value of the controlling expression is 0, or "false", and else is omitted, then the program execution continues with the next statement.

If several if statements are nested, then an else clause always belongs to the last if (in the given block nesting level) that does not yet have an else clause. An else can be assigned to a different if by creating explicit blocks.

Example:

if ( n > 0 
{  if ( n % 2 == 0 
     puts("n is positive and even");
}
else                        // Belongs to first if 
     puts("n is negative or zero");

Name

switch

Synopsis

In a switch statement, the value of the switch expression is compared to the constants associated with case labels. If the expression evaluates to the constant associated with a case label, program execution continues at the matching label. If no matching label is present, program execution branches to the default label if present; otherwise execution continues with the statement following the switch statement.

Syntax:

switch ( 
                     expression ) 
                     statement

The expression is an integer expression and statement is a block statement with case labels and at most one default label. Every case label has the form case const :, where const is a constant integer expression. All case constants must be different from one another.

Example:

switch( command )          // Query a command obtained
{                          // by user input in a menu, 
                           // for example.
  case 'a':
  case 'A': action1();     // Carry out action 1,
            break;         // then quit the switch.
  case 'b':
  case 'B': action2();     // Carry out action 2,
            break;         // then quit the switch.
  default:  putchar('\a'); // On any other "command":
                           // alert.
}

After the jump from the switch to a label, program execution continues sequentially, regardless of other labels. The break statement can be used to exit the switch block at any time. A break is thus necessary if the statements following other case labels are not to be executed.

Integer promotion is applied to the switch expression. The case constants are then converted to the resulting type of the switch expression.

Name

while

Synopsis

The while statement is a "top-driven" loop: first the loop condition (i. e., the controlling expression) is evaluated. If it yields "true", the loop body is executed, and then the controlling expression is evaluated again. If it is false, program execution continues with the statement following the loop body.

Syntax:

while ( expression )  statement

Example:

s = str;           // Let the char pointer s  
while( *s != '\0') // point to the end of str 
    ++s;

Name

do ... while

Synopsis

The do ... while statement is a "bottom-driven" loop: first the body of the loop is executed, then the controlling expression is evaluated. This is repeated until the controlling expression is "false", or 0.

The key difference from a while statement is that a do ... while loop body is always executed at least once. A while loop may not execute at all, because its expression could be false to begin with.

Syntax:

do statement  while ( expression ) ;

Example:

i = 0;
do                      // Copy the string str1
    str2[i] = str1[i];  // to string str2 
while ( str1[i++] != '\0' );

Name

for

Synopsis

A typical for loop uses a control variable and performs the following actions on it:

1. Initialization (once before beginning the loop)

2. Tests the controlling expression

3. Makes adjustments (such as incrementation) at the end of each loop iteration

The three expressions in the head of the for loop define these three actions.

Syntax:

for ([expression1]; [expression2]; [expression3]
         
                     statement

expression1 and expression3 can be any expressions. Expression2 is the controlling expression, and hence must have a scalar type. Any of these expressions can be omitted. If expression2 is omitted, the loop body is executed unconditionally. In ANSI C99, expression1 may also be a declaration. The scope of the variable declared is then limited to the for loop.

Example:

for (int i = DELAY; i > 0; --i)    // Wait a little
    ;

Except for the scope of the variable i, this for loop is equivalent to the following while loop:

int i = DELAY; // Initialize 
while( i > 0)  // Test the controlling expression

     --i; // Adjust

Name

goto

Synopsis

The goto statement jumps to any point within a function. The destination of the jump is specified by the name of a label.

Syntax:

goto label_name;

A label is a name followed by a colon that appears before any statement.

Example:

for ( ... )         // Jump out of
    for ( ... )     // nested loops. 
       if ( error 
         goto  handle_error;
  ...
 handle_error:      // Error handling here
 ...

The only restriction is that the goto statement and the label must be contained in the same function. Nonetheless, the goto statement should never be used to jump into a block from outside it.

Name

continue

Synopsis

The continue statement can only be used within the body of a loop. It jumps over the remainder of the loop body. Thus in a while or do ... while loop, it jumps to the next test of the controlling expression, and in a for loop it jumps to the evaluation of the per-iteration adjustment expression.

Syntax:

continue;

Example:

for (i = -10; i < 10; ++i
{  ...
  if (i == 0) continue;     // Skip the value 0
   ...
}

Name

break

Synopsis

The break statement jumps immediately to the statement after the end of a loop or switch statement. This provides a way to end execution of a loop at any point in the loop body.

Syntax:

break;

Example:

while (1
{  ...
   if (command == ESC) break;     // Exit the loop 
   ...  
}

Name

return

Synopsis

The return statement ends the execution of the current function and returns control to the caller. The value of the expression in the return statement is returned to the caller as the return value of the function.

Syntax:

return expression;

Example:

int max( int a, int b )     // The maximum of a and b
{ return (a>b ? a : b); }

Any number of return statements can appear in a function.

The value of the return expression is converted to the type of the function if necessary.

The expression in the return statement can be omitted. This only makes sense in functions of type void, however—in which case the entire return statement can also be omitted. Then the function returns control to the caller at the end of the function block.

Declarations

A declaration determines the interpretation and properties of one or more identifiers. A declaration that allocates storage space for an object or a function is a definition. In C, an object is a data storage region that contains constant or variable values. The term "object" is thus somewhat more general than the term "variable."

In the source file, declarations can appear at the beginning of a block, such as a function block, or outside of all functions. Declarations that do not allocate storage space, such as function prototypes or type definitions, are normally placed in a header file.

ANSI C99 allows declarations and statements to appear in any order within a block.

General Syntax and Examples

The general syntax of a declaration is as follows:

[storage class] type  D1 [, D2, ...];

storage class

One of the storage class specifiers extern, static, auto, or register.

type

A basic type, or one of the following type specifiers: void, enum type (enumeration), struct or union type, or typedef name.

type may also contain type qualifiers, such as const.

D1 [, D2,...]

A list of declarators. A declarator contains at least one identifier, such as a variable name.

Some examples are:

char  letter;
int   i, j, k;
static  double rate, price;
extern char  flag;

Variables can be initialized—that is, assigned an initial value—in the declaration. Variable and function declarations are described in detail in the sections that follow.

Complex Declarations

If a declarator contains only one identifier, with or without an initialization, the declaration is called a simple declaration. In a complex declaration, the declarator also contains additional type information. This is necessary in declarations of pointers, arrays, and functions. Such declarations use the three operators, shown in Table 1-17.

Table 1-17. Operators for complex declarations

The operators in Table 1-17 have the same precedence in declarations as in expressions. Parentheses can also be used to group operands.

Complex declarators are always interpreted beginning with the identifier being declared. Then the following steps are repeated until all operators are resolved:

1. Any pair of parentheses () or square brackets [] appearing to the right is interpreted.

2. If there are none, then any asterisk appearing to the left is interpreted.

For example:

char *strptr[100];

This declaration identifies strptr as an array. The array's 100 elements are pointers to char.

Variables

Every variable must be declared before it can be used. The declaration determines the variable's type, its storage class, and possibly its initial value. The type of a variable determines how much space it occupies in storage and how the bit pattern it stores is interpreted. For example:

float dollars = 2.5F;     // a variable of type float

The variable dollars designates a region in memory with a size of 4 bytes. The contents of these four bytes are interpreted as a floating-point number, and initialized with the value 2.5.

Storage Classes

The storage class of a variable determines its scope, its storage duration, and its linkage. The scope can be either block or file (see Section 1.2.4, earlier in this book). Variables also have one of two storage durations:

Static storage duration

The variable is generated and initialized once, before the program begins. It exists continuously throughout the execution of the program.

Automatic storage duration

The variable is generated anew each time the program flow enters the block in which it is defined. When the block is terminated, the memory occupied by the variable is freed.

The storage class of a variable is determined by the position of its declaration in the source file and by the storage class specifier, if any. A declaration may contain no more than one storage class specifier. Table 1-18 lists the valid storage class specifiers.

Table 1-18. The storage class specifiers

Table 1-19 illustrates the possible storage classes and their effect on the scope and the storage duration of variables.

Table 1-19. Storage class, scope, and storage duration of variables

Initialization

Variables can be initialized (assigned an initial value) in their declaration. The initializer consists of an equal sign followed by a constant expression. Some examples are:

int index = 0, max = 99, *intptr = NULL;
static char message[20] = "Example!";

Variables are not initialized in declarations that do not cause an object to be created, such as function prototypes and declarations that refer to external variable definitions.

Every initialization is subject to the following rules:

1. A variable declaration with an initializer is always a definition. This means that storage is allocated for the variable.

2. A variable with static storage duration can only be initialized with a value that can be calculated at the time of compiling. Hence the initial value must be a constant expression.

3. For declarations without an initializer: variables with static storage duration are implicitly initialized with NULL (all bytes have the value 0); the initial value of all other variables is undefined!

The type conversion rules for simple assignments are also applied on initialization.

Derived Types

A programmer can also define new types, including enumerated types and derived types. Derived types include pointers, arrays, structures, and unions.

The basic types and the enumerated types are collectively called the arithmetic types . The arithmetic types and the pointer types in turn make up the scalar types . The array and structure types are known collectively as the aggregate types .

Enumeration Types

Enumeration types are used to define variables that can only be assigned certain discrete integer values throughout the program. The possible values and names for them are defined in an enumeration. The type specifier begins with the keyword enum ; for example:

enum toggle { OFF, ON, NO = 0, YES };

The list of enumerators inside the braces defines the new enumeration type. The identifier toggle is the tag of this enumeration. This enumeration defines the identifiers in the list (OFF, ON, NO, and YES) as constants with type int.

The value of each identifier in the list may be determined explicitly, as in NO = 0 in the example above. Identifiers for which no explicit value is specified are assigned a value automatically based on their position in the list, as follows: An enumerator without an explicit value has the value 0 if it is the first in the list; otherwise its value is 1 greater than that of the preceding enumerator. Thus in the example above, the constants OFF and NO have the value 0, while ON and YES have the value 1.

Once an enumeration type has been defined, variables with the type can be declared within its scope. For example:

enum toggle t1, t2 = ON;

This declaration defines t1 and t2 as variables with type enum toggle, and also initializes t2 with the value ON, or 1.

Following is an enumeration without a tag:

enum { black, blue, green, cyan, red, magenta, white };

As this example illustrates, the definition of an enumeration does not necessarily include a tag. In this case, the enumeration type cannot be used to declare variables, but the enumeration constants can be used to designate a set of discrete values. This technique can be used as an alternative to the #define directive. The constants in the example above have the following values: black = 0, blue = 1, ... , white = 6.

Variables with an enumeration type can generally be used in a C program—in comparative or arithmetic expressions, for example—as ordinary int variables.

Structures, Unions, and Bit-Fields

Different data items that make up a logical unit are generally grouped together in a record. The structure of a record—i. e., the names, types, and order of its components—is represented in C by a structure type .

The components of a record are called the members of the structure. Each member can be of any type. The type specifier begins with the keyword struct ; for example:

struct article    {    char   name[40];
                       int    quantity;
                       double price;
                  };

This example declares a structure type with three members. The identifier article is the tag of the structure, and name, quantity, and price are the names of its members. Within the scope of a structure declaration, variables can be declared with the structure type:

struct article  a1, a2, *pArticle, arrArticle[100];

a1 and a2 are variables of type struct article, and pArticle is a pointer to an object of type struct article. The array arrArticle has 100 elements of type struct article.

Structure variables can also be declared simultaneously with the structure type definition. If no further reference is made to a structure type, then its declaration need not include a tag. For example:

struct {unsigned char character, attribute;}
xchar, xstr[100];

The structure type defined here has the members character and attribute, both of which have the type unsigned char. The variable xchar and the elements of the array xstr have the type of the new tagless structure.

The members of a structure variable are located in memory in order of their declaration within the structure. The address of the first member is identical to the address of the entire structure. The addresses of the other members and the total storage space required by the structure may vary, however, since the compiler can insert unnamed gaps between the individual members for the sake of optimization. For this reason the storage size of a structure should always be obtained using the sizeof operator.

The macro offsetof , defined in the header file stddef.h , can be used to obtain the location of a member within a structure. The expression:

offsetof( structure_type, member )

has the type size_t, and yields the distance in bytes between the beginning of the structure and member.

Structure variables can be initialized by an initialization list containing a value for each member:

struct article flower =     // Declare and initialize the
  { "rose", 7, 2.49 };      // structure variable flower

A structure variable with automatic storage duration can also be initialized with the value of an existing structure variable. The assignment operator can be used on variables of the same structure type. For example:

arrArticle[0] = flower;

This operation copies the value of each member of flower to the corresponding member of arrArticle[0].

A specific structure member can be accessed by means of the dot operator, which has a structure variable and the name of a member as its operands:

 flower.name    // The array 'name' 
 flower.price   // The double variable 'price'

Efficient data handling often requires the use of pointers to structures . The arrow operator provides convenient access to a member of a structure identified by a pointer. The left operand of the arrow operator is a pointer to a structure. Some examples follow:

pArticle = &flower;     // Let pArticle point to flower
pArticle->quantity      // Access members of flower
pArticle->price         // using the pointer pArticle

A structure cannot have itself as a member. Recursive structures can be defined, however, by means of members that are pointers to the structure's own type. Such recursive structures are used to implement linked lists and binary trees, for example.

Unions

A union permits references to the same location in memory to have different types. The declaration of a union differs from that of a structure only in the keyword union :

union number {long n; double x;};

This declaration creates a new union type with the tag number and the two members n and x.

Unlike the members of a structure, all the members of a union begin at the same address! Hence the size of a union is that of its largest member. According to the example above, a variable of type union number occupies 8 bytes.

Once a union type has been defined, variables of that type can be declared. Thus:

union number  nx[10];

declares an array nx with ten elements of type union number. At any given time, each such element contains either a long or a double value. The members of a union can be accessed in the same ways as structure members. For example:

nx[0].x = 1.234;   // Assign a double value to nx[0]

Like structures, union variables are initialized by an initializer list. For a union, however, the list contains only one initializer. If no union member is explicitly designated, the first member named in the union type declaration is initialized:

union number length = { 100L };

After this declaration, length.n has the value 100.

Bit-fields

Members of structures or unions can also be bit-fields. Bit-fields are integers which consist of a defined number of bits. The declaration of a bit-field has the form:

                  type    [identifier] : width;

where type is either unsigned int or signed int, identifier is the optional name of the bit-field, and width is the number of bits occupied by the bit-field in memory.

A bit-field is normally stored in a machine word that is a storage unit of length sizeof(int). The width of a bit-field cannot be greater than that of a machine word. If a smaller bit-field leaves sufficient room, subsequent bit-fields may be packed into the same storage unit. A bit-field with width zero is a special case, and indicates that the subsequent bit-field is to be stored in a new storage unit regardless of whether there's room in the current storage unit. Here's an example of a structure made up of bit fields:

    struct     {    unsigned int  b0_2 : 3;
                    signed   int  b3_7 : 5;
                    unsigned int       : 7;
                    unsigned int  b15  : 1;
               } var;

The structure variable var occupies at least two bytes, or 16 bits. It is divided into four bit-fields: var.b0_2 occupies the lowest three bits, var.b3_7 occupies the next five bits, and var.b15 occupies the highest bit. The third member has no name, and only serves to define a gap of seven bits, as shown in Figure 1-5.

Figure 1-5. Bit assignments in the example struct

Bit-fields with the type unsigned int are interpreted as unsigned. Bit-fields of type signed int can have negative values in two's-complement encoding. In the example above, var.b0_2 can hold values in the range from 0 to 7, and var.b3_7 can take values in the range from -16 to 15.

Bit-fields also differ from ordinary integer variables in the following ways:

• The address operator (&) cannot be applied to bit-fields (but it can be applied to a structure variable that contains bit-fields).

• Some uses of bit-fields may lead to portability problems, since the interpretation of the bits within a word can differ from one machine to another.

Arrays

Arrays are used to manage large numbers of objects of the same type. Arrays in C can have elements of any type except a function type. The definition of an array specifies the array name, the type, and, optionally, the number of array elements. For example:

    char line[81];

The array line consists of 81 elements with the type char. The variable line itself has the derived type "array of char" (or "char array").

In a statically defined array, the number of array elements (i. e., the length of the array) must be a constant expression. In ANSI C99, any integer expression with a positive value can be used to specify the length of a non-static array with block scope. This is also referred to as a variable-length array .

An array always occupies a continuous location in memory. The size of an array is thus the number of elements times the size of the element type:

sizeof( line ) == 81 * sizeof( char ) == 81 bytes

The individual array elements can be accessed using an index. In C, the first element of an array has the index 0. Thus the 81 elements of the array line are line[0], line[1], ... , line[80].

Any integer expression can be used as an index. It is up to the programmer to ensure that the value of the index lies within the valid range for the given array.

A string is a sequence of consecutive elements of type char that ends with the null character, '\0'. The length of the string is the number of characters excluding the string terminator '\0'. A string is stored in a char array, which must be at least one byte longer than the string.

A wide string consists of characters of type wchar_t and is terminated by the wide null character, L'\0'. The length of a wide string is the number of wchar_t characters in the string, excluding the wide string terminator. For example:

    wchar_t wstr[20] = L"Mister Fang"; //  length: 11
                                       //  wide characters

A multi-dimensional array in C is an array whose elements are themselves arrays. For example:

    short point[50][20][10];

The three-dimensional array point consists of 50 elements that are two-dimensional arrays. The declaration above defines a total of 50*20*10 = 10,000 elements of type short, each of which is uniquely identified by three indices:

point[0][0][9] = 7;  // Assign the value 7 to  the "point" 
                     // with the "coordinates" (0,0,9).

Two-dimensional arrays, also called matrices , are the most common multi-dimensional arrays. The elements of a matrix can be thought of as being arranged in rows (first index) and columns (second index).

Arrays in C are closely related to pointers: in almost all expressions, the name of an array is converted to a pointer to the first element of the array. The sizeof operator is an exception, however: if its operand is an array, it yields the number of bytes occupied, not by a pointer, but by the array itself. After the declaration:

char msg[] = "Hello, world!";

the array name msg points to the character 'H'. In other words, msg is equivalent to &msg[0]. Thus in a statement such as:

puts( msg ); // Print string to display

only the address of the beginning of the string is passed to the function puts(). Internally, the function processes the characters in the string until it encounters the terminator character '\0'.

An array is initialized by an initialization list containing a constant initial value for each of the individual array elements:

double x[3] = { 0.0, 0.5, 1.0 };

After this definition, x[0] has the value 0.0, x[1] the value 0.5, and x[2] the value 1.0. If the length of the array is greater than the number of values in the list, then all remaining array elements are initialized with 0. If the initialization list is longer than the array, the redundant values are ignored.

The length of the array need not be explicitly specified, however:

double x[] = { 0.0, 0.5, 1.0 };

In this definition, the length of the array is determined by the number of values in the initialization list.

A char array can be initialized by a string literal:

char str[] = "abc";

This definition allocates and initializes an array of four bytes, and is equivalent to:

char str[] = { 'a', 'b', 'c', '\0' } ;

In the initialization of a multi-dimensional array , the magnitude of all dimensions except the first must be specified. In the case of a two-dimensional array, for example, the number of rows can be omitted. For example:

char error_msg[][40] =    { "Error opening file!",
                            "Error reading file!",
                            "Error writing to file!"};

The array error_msg consists of three rows, each of which contains a string.

Pointers

A pointer represents the address and type of a variable or a function. In other words, for a variable x, &x is a pointer to x.

A pointer refers to a location in memory, and its type indicates how the data at this location is to be interpreted. Thus the pointer types are called pointer to char, pointer to int, and so on, or for short, char pointer, int pointer, etc.

Array names and expressions such as &x are address constants or constant pointers, and cannot be changed. Pointer variables, on the other hand, store the address of the object to which they refer, which address you may change. A pointer variable is declared by an asterisk (*) prefixed to the identifier. For example:

float  x, y, *pFloat;
pFloat = &x;     // Let pFloat point to x.

After this declaration, x and y are variables of type float, and pFloat is a variable of type float * (pronounced "pointer to float"). After the assignment operation, the value of pFloat is the address of x.

The indirection operator * is used to access data by means of pointers. If ptr is a pointer, for example, then *ptr is the object to which ptr points. For example:

y = *pFloat;   //  equivalent to  y = x;

As long as pFloat points to x, the expression *pFloat can be used in place of the variable x. Of course, the indirection operator * must only be used with a pointer which contains a valid address.

A pointer with the value 0 is called a null pointer . Null pointers have a special significance in C. Because all objects and functions have non-zero addresses, a null pointer always represents an invalid address. Functions that return a pointer can therefore return a null pointer to indicate a failure condition. The constant NULL is defined in stdio.h, stddef.h, and other header files as a null pointer (i.e., a pointer with a value of zero).

All object pointer variables have the same storage size, regardless of their type. Two or four bytes are usually required to store an address.

Parentheses are sometimes necessary in complex pointer declarations. For example:

long arr[10];     // Array arr with ten elements
long (*pArr)[10]; // Pointer pArr to an array
                  //  of ten long elements

Without the parentheses, the declaration long *pArr[10]; would create an array of ten pointers to long. Parentheses are always necessary in order to declare pointers to arrays or functions.

Pointer arithmetic

Two arithmetic operations can be performed on pointers:

• An integer can be added to or subtracted from a pointer.

• One pointer can be subtracted from another of the same type.

These operations are generally useful only when the pointers point to elements of the same array. In arithmetic operations on pointers, the size of the objects pointed to is automatically taken into account. For example:

int a[3] = { 0, 10, 20 };  // An array with three elements
int *pa = a;               // Let pa point to a[0]

Since pa points to a[0], the expression pa + 1 yields a pointer to the next array element, a[1], which is sizeof( int ) bytes away in memory. Furthermore, because the array name a likewise points to a[0], a+1 also yields a pointer to a[1].

Thus for any integer i, the following expressions are equivalent:

&a[i] , a+i , pa+i // pointers to the i-th array element

By the same token, the following expressions are also equivalent:

a[i] , *(a+i) , *(pa+i) , pa[i]  // the i-th array element

Thus a pointer can be treated as an array name: pa[i] and *(pa+i) are equivalent. Unlike the array name, however, pa is a variable, not an address constant. For example:

pa = a+2;      // Let pa point to a[2]
int n = pa-a;  //  n = 2

The subtraction of two pointers yields the number of array elements between the pointers. For example, the expression pa-a yields the integer value 2 if pa points to a[2]. This value has the integer type ptrdiff_t, which is defined (usually as int) in stddef.h.

The addition of two pointers is not a useful operation, and hence is not permitted. It is possible, however, to compare two pointers of the same type, as the following example illustrates:

// Formatted output of the elements of an array
#define  LEN  10
float numbers[LEN], *pn;
     . . . 
for ( pn = numbers; pn < numbers+LEN; ++pn ) 
   printf( "%16.4f", *pn );

Function pointers

The name of a function is a constant pointer to the function. Its value is the address of the function's machine code in memory. For example, the name puts is a pointer to the function puts(), which outputs a string:

#include <stdio.h>        // Include declaration of puts()
int (*pFunc)(const char*);  // Pointer to a function 
    . . .                   // whose parameter is a string
                            // and whose return value 
                            // has type int
pFunc = puts;               // Let pFunc point to puts()
(*pFunc)("Any questions?"); // Call puts() using the
                            // pointer

Note that the first pair of parentheses is required in the declaration of the variable pFunc. Without it, int *pFunc( const char* ); would declare pFunc as a function that returns a pointer to int.

Type Qualifiers and Type Definitions

The type of an object can be qualified by the keywords const and volatile in the declaration.

The type qualifier const indicates that the program can no longer modify an object after its declaration. For example:

const double pi = 3.1415927;

After this declaration, a statement that modifies the object pi, such as pi = pi+1;, is illegal and results in a compiler error.

The type qualifier volatile indicates variables that can be modified by processes other than the present program. Based on this information, the compiler may refrain from optimizing access to the variable.

The type qualifiers volatile and const can also be combined:

extern const volatile unsigned  clock_ticks;

After this declaration, clock_ticks cannot be modified by the program, but may be modified by another process, such as a hardware clock interrupt handler.

Type qualifiers are generally prefixed to the type specifier. In pointer declarations, however, type qualifiers may be applied both to the pointer itself and to the object it addresses. If the type qualifier is to be applied to the pointer itself, it must be placed immediately before the identifier.

The most common example of such a declaration is the "pointer to a constant object." Such a pointer may point to a variable, but cannot be used to modify it. For this reason, such pointers are also called "read-only" pointers. For example:

int var1 = 1, var2 = 2, *ptr; 
const int cArr[2];
const int *ptrToConst;// "Read-only pointer" to int

The following statements are now permitted:

ptrToConst = &cArr[0];    // Change the value of 
++ptrToConst;             // the pointer variable
ptrToConst = &var1;
var2 = *ptrToConst;       // "Read" access

The following statements are not permitted:

ptr = ptrToConst;    // "Read-only" cannot be copied to
                     // "read-write"
*ptrToConst = 5;     // "Write" access not allowed!

Name

restrict

Synopsis

ANSI C99 introduces the type qualifier restrict , which is only applicable to pointers. If a pointer declared with the restrict qualifier points to an object that is to be modified, then the object can only be accessed using that pointer. This information allows the compiler to generate optimized machine code. It is up to the programmer to ensure that the restriction is respected!

Example:

void *memcpy( void * restrict dest,     // destination
              const void* restrict src, // source
              size_t n );

In using the standard function memcpy() to copy a memory block of n bytes, the programmer must ensure that the source and destination blocks do not overlap.

Name

typedef

Synopsis

The keyword typedef is used to give a type a new name.

Examples:

typedef unsigned char UCHAR;
typedef struct { double x, y } POINT;

After these type definitions, the identifier UCHAR can be used as an abbreviation for the type unsigned char, and the identifier POINT can be used to specify the given structure type.

Examples:

UCHAR  c1, c2, tab[100];
POINT point, *pPoint;

In a typedef declaration, the identifier is declared as the new type name. The same declaration without the typedef keyword would declare a variable and not a type name.

Functions

Every C program contains at least the function main() , which is the first function executed when the program starts. All other functions are subroutines.

The definition of a function lists the statements it executes. Before a function can be called in a given translation unit, it must be declared. A function definition also serves as a declaration of the function. The declaration of a function informs the compiler of its return type. For example:

extern double pow();

Here pow() is declared as a function that returns a value with type double. Because function names are external names by default, the storage class specifier extern can also be omitted.

In ANSI C99, implicit function declarations are no longer permitted. Formerly, calls to undeclared functions were allowed, and the compiler implicitly assumed in such cases that the function returned a value of type int.

The declaration of the function pow() in the example above contains no information about the number and type of the function's parameters. Hence the compiler has no way of testing whether the arguments supplied in a given function call are compatible with the function's parameters. This missing information is supplied by a function prototype.

Function Prototypes

A function prototype is a declaration that indicates the types of the function's parameters as well as its return value. For example:

double pow( double, double );    // prototype of pow()

This prototype informs the compiler that the function pow() expects two arguments of type double, and returns a result of type double. Each parameter type may be followed by a parameter name. This name has no more significance than a comment, however, since its scope is limited to the function prototype itself. For example:

double pow( double base, double exponent );

Functions that do not return any result are declared with the type specifier void. For example:

void func1( char *str ); // func1 expects one string
                         // argument and has no return 
                         // value.

Functions with no parameters are declared with the type specifier void in the parameter list:

int  func2( void );     // func2 takes no arguments and
                        // returns a value with type int.

Function declarations should always be in prototype form. All standard C functions are declared in one (or more) of the standard header files. For example, math.h contains the prototypes of the mathematical functions, such as sin(), cos(), pow(), etc., while stdio.h contains the prototypes of the standard input and output functions.

Function Definitions

The general form of a function definition is:

[storage_class] [type] name( 
[parameter_list] ) // function declarator
{
    /* declarations, statements */      // function body
}

storage_class

One of the storage class specifiers extern or static. Because extern is the default storage class for functions, most function definitions do not include a storage class specifier.

type

The type of the function's return value. This can be either void or any other type, except an array.

name

The name of the function.

parameter_list

The declarations of the function's parameters. If the function has no parameters, the list is empty.

Here is one example of a function definition:

long sum( int arr[], int len )// Find the sum of the first 
{                         // len elements of the array arr
  int i;
  long result = 0;       

  for( i = 0;  i < len;  ++i )
     result += (long)arr[i];
  return result;
}

Because by default function names are external names, the functions of a program can be distributed among different source files, and can appear in any sequence within a source file.

Functions that are declared as static, however, can only be called in the same translation unit in which they are defined. But it is not possible to define functions with block scope—in other words, a function definition cannot appear within another function.

The parameters of a function are ordinary variables whose scope is limited to the function. When the function is called, they are initialized with the values of the arguments received from the caller.

The statements in the function body define what the function does. When the flow of execution reaches a return statement or the end of the function body, control returns to the calling function.

A function that calls itself, directly or indirectly, is called recursive . C permits the definition of recursive functions, since variables with automatic storage class are created anew—generally in stack memory—with each function call.

The function declarator shown above is in prototype style. Today's compilers still support the older Kernighan-Ritchie style, however, in which the parameter identifiers and the parameter type declarations are separate. For example:

long sum( arr, len )  // Parameter identifier list
int arr[], len;       // Parameter declarations 
{ ... }               // Function body

In ANSI C99, functions can also be defined as inline. The inline function specifier instructs the compiler to optimize the speed of the function call, generally by inserting the function's machine code directly into the calling routine. The inline keyword is prefixed to the definition of the function:

inline int max( int x, int y )
{  return  ( x >= y ? x : y ); }

If an inline function contains too many statements, the compiler may ignore the inline specifier and generate a normal function call.

An inline function must be defined in the same translation unit in which it is called. In other words, the function body must be visible when the inline "call" is compiled. It is therefore a good idea to define inline functions—unlike ordinary functions—in a header file.

Inline functions are an alternative to macros with parameters. In translating a macro, the preprocessor simply substitutes text. An inline function, however, behaves like a normal function—so that the compiler tests for compatible arguments, for example—but without the jump to and from another code location.

Function Calls

A function call is an expression whose value and type are those of the function's return value.

The number and the type of the arguments in a function call must agree with the number and type of the parameters in the function definition. Any expression, including constants and arithmetic expressions, may be specified as an argument in a function call. When the function is called, the value of the argument is copied to the corresponding parameter of the function! For example:

double x=0.5, y, pow();    // Declaration
y = pow( 1.0 + x, 2.5 );   // Call to pow() yields 
                           // the double value (1.0+x)^2.5

In other words, the arguments are passed to the function by value. The function itself cannot modify the values of the arguments in the calling function: it can only access its local copy of the values.

In order for a function to modify the value of a variable directly, the caller must give the function the address of the variable as an argument. In other words, the variable must be passed to the function by reference. Examples of functions that accept arguments by reference include scanf(), time(), and all functions that have an array as one of their parameters. For example:

double swap( double *px, double *py ) // Exchange values
                                      // of two variables
{ double z = *px; *px = *py; *py = z; }

The arguments of a function are subject to implicit type conversion:

• If the function was declared in prototype form (as is usually the case), each argument is converted to the type of the corresponding parameter, as for an assignment.

• If no prototype is present, integer promotion is performed on each integer argument. Arguments of type float are converted to double.

Functions with Variable Numbers of Arguments

Functions that can be called with a variable number of arguments always expect a fixed number of mandatory arguments—at least one is required—and a variable number of optional arguments. A well-known example is the function printf(): the format string argument is mandatory, while all other arguments are optional. Internally, printf() determines the number and type of the other arguments from the information in the format string.

In the function declarator, optional arguments are indicated by three dots (...). For example:

int printf( char *str, ... );     // Prototype

In the function definition, the optional arguments are accessed through an object with the type va_list , which contains the argument information. This type is defined in the header file stdarg.h , along with the macros va_start , va_arg , and va_end , which are used to manage the arguments.

In order to read the optional arguments, the function must carry out the following steps:

1. Declare an object of type va_list. In the following example, this object is named arglist.

2. Invoke the macro va_start to prepare the arglist object to return the first optional argument. The parameters of va_start are the arglist object and the name of the last mandatory parameter.

3. Invoke the macro va_arg with the initialized arglist object to obtain each of the optional arguments in sequence. The second parameter of va_arg is the type of the optional argument that is being obtained.

   After each invocation of the va_arg macro, the arglist object is prepared to deliver the first optional argument that has not yet been read. The result of va_arg has the type specified by the second argument.

4. After reading out the argument list, the function should invoke the va_end macro before returning control to the caller. The only parameter of va_end is the arglist object.

Following is an example of a function, named max, that accepts a variable number of arguments:

// Determine the maximum of a number of positive integers. 
// Parameters: a variable number of positive values of
// type unsigned int. The last argument must be 0.
// Return value: the maximum of the arguments 

#include <stdarg.h>
unsigned int max( unsigned int first, ... )
{
   unsigned int maxarg, arg;
   va_list arglist;  // The optional-argument 
                     // list object
   va_start( arglist, first ); // Set arglist to deliver
                               // the first optional
                               // argument  
    arg = maxarg = first;
    while ( arg != 0 )
    { arg = va_arg( arglist, unsigned );// Get an argument
       if ( arg > maxarg )  maxarg = arg;
    }
    va_end( arglist );         // Finished reading the 
                               // optional arguments 
    return maxarg; 
}

Linkage of Identifiers

An identifier that is declared more than once, whether in different scopes (in different files, for example) or in the same scope, may refer to the same variable or function. Identifiers must be "linked" in this way in order for a variable to be used "globally," across different source files, for example.

Each identifier has either external, internal, or no linkage. These three kinds of linkage have the following significance:

External linkage

An identifier with external linkage represents the same object or function throughout the entire program, i. e., in all source files and libraries belonging to the program. The identifier is made known to the linker.

When a second declaration of the same identifier with external linkage occurs, the linker associates the identifier with the same object or function. A declaration of an existing external object is sometimes called a reference declaration .

Internal linkage

An identifier with internal linkage represents the same object or function within a given translation unit. The linker has no information about identifiers with internal linkage. Thus they remain "internal" to the translation unit.

No linkage

If an identifier has no linkage, then any further declaration using the identifier declares something new, such as a new variable or a new type.

The linkage of an identifier is determined by its storage class; that is, by the position of the declaration and any storage class specifier included in it. Only identifiers of variables and functions can have internal or external linkage. All other identifiers, and identifiers of variables with automatic storage class, have no linkage. Table 1-20 summarizes this information.

Table 1-20. Linkage of identifiers

The form of external names (identifiers with external linkage) is subject to restrictions, depending on the linker implementation: some linkers only recognize the first eight characters of a name, and do not distinguish between upper- and lower-case letters.

Preprocessing Directives

The C compiler preprocesses every source file before performing the actual translation. The preprocessor removes comments and replaces macros with their definitions.

Every preprocessing directive appears on a line by itself, beginning with the character #. If the directive is long, it can be continued on the next line by inserting a backslash (\) as the last character before the line break.

Name

#define

Synopsis

The #define directive is used to define macros.

Syntax:

define name[(parameter_list)]  [replacement_text]

The preprocessor replaces each occurrence of name or name(parameter_list) in the subsequent source code with replacement_text.

Examples:

#define  BUF_SIZE  512        // Symbolic constant
#define  MAX(a,b)  ((a) > (b) ? (a) : (b))

These directives define the macros BUF_SIZE and MAX. If the replacement text is a constant expression, the macro is also called a symbolic constant . Macros can also be nested; a macro, once defined, can be used in another macro definition.

In the previous example, the parentheses are necessary in order for the substitution to be performed correctly when MAX is used in an expression, or when complex expressions replace the parameters a and b. For example, the preprocessor replaces the macro invocation:

result = 2 * MAX( x, y & 0xFF );

with:

result = 2 * ( (x) > (y & 0xFF) ? (x) : (y & 0xFF) );

Name

The # Operator

Synopsis

In the macro replacement text, the parameters of the macro may be preceded by the operator # (called the hash or stringizing operator). In this case, the preprocessor sets the corresponding argument in quotation marks, thus converting it into a string.

Example:

#define print_int(i) printf( "value " #i " = %d", i )

If x and y are variables with type int, then the statement:

print_int(x-y);

is replaced with:

printf( "value ""x-y"" = %d", x-y );

Because consecutive string literals are concatenated, this is equivalent to:

printf( "value x-y = %d", x-y );

Name

The ## Operator

Synopsis

If a macro parameter appears in the replacement text preceded or followed by the operator ## (called the double-hash or token-pasting operator), then the preprocessor concatenates the tokens to the left and right of the operator, ignoring any spaces. If the resulting text also contains a macro name, then macro replacement is performed once again.

Example:

#define show( var, num )  \
       printf( #var #num " = %.1f\n", var ## num )

If the float variable x5 has the value 16.4, then the macro invocation:

show( x, 5 );

is replaced with:

printf( "x" "5" " = %.1f\n", x5 ); 
// Output: x5 = 16.4\n

Name

#undef

Synopsis

The #undef directive cancels a macro definition. This is necessary when the definition of a macro needs to be changed, or when a function of the same name needs to be called.

Syntax:

 #undef name

No parameter list needs to be specified, even if the previously defined macro has parameters.

Example:

#include <ctype.h>
#undef  toupper
   . . . 
c = toupper(c);  // Call the function toupper()

Name

#include

Synopsis

The #include directive instructs the preprocessor to insert the contents of a specified file in the program at the point where the #include directive appears.

Syntax:

#include  <filename>
#include  "filename"

If the filename is enclosed in angle brackets, the preprocessor only searches for it in certain directories. These directories are usually named in the environment variable INCLUDE .

If the filename is enclosed in quotation marks, the preprocessor first looks for the file in the current working directory.

The filename may contain a directory path. In this case, the file is only looked for in the specified directory.

The files named in include directives are generally "header" files containing declarations and macro definitions for use in several source files, and have names ending in .h. Such files may in turn contain further #include directives.

In the following example, one file to be included is selected based on the value of a symbolic constant:

#include <stdio.h>
#include "project.h"

#if VERSION == 1
  #define  MYPROJ_H  "version1.h"
#else
  #define  MYPROJ_H  "version2.h"
#endif

#include MYPROJ_H

Name

#if, #elif, #else, #endif

Synopsis

These directives are used to present source code to the compiler only on certain conditions. In this way a different selection of program statements can be compiled from one build to another. This technique can be used to adapt a single program to a variety of target systems, for example, without requiring modification of the source code.

Syntax:

#if expression1
  [text1]
[#elif  expression2
                   text2]
     
    . . .
[#elif  expression(n)
                   text(n)]
[#else
  text(n+1)]
 #endif

Each #if directive may be followed by any number of #elif directives, and at most one #else directive. The conditional source code section must be closed by an #endif directive.

The preprocessor evaluates expression1, expression2, etc. in succession. At the first expression whose value is "true", i. e., not equal to 0, the conditional code is processed. If none of the expressions is true, then the #else directive is processed, if present.

expression1, expression2, etc. must be constant integer expressions. The cast operator cannot be used in preprocessing directives.

The conditional text consists of program code, including other preprocessing directives and ordinary C statements. Conditional text that the preprocessor skips over is effectively removed from the program.

Name

The defined operator

Synopsis

The defined operator can be used to verify whether a given macro name is currently defined.

Syntax:

defined (name)

The operator yields a non-zero value if a valid definition exists for name; otherwise it yields the value 0. A macro name defined by a #define directive remains defined until it is cancelled by an #undef directive. A macro name is considered to be defined even if no replacement text is specified after name in the #define directive.

The defined operator is typically used in #if and #elif directives:

#if defined(VERSION
...
#endif

Unlike the #ifdef and #ifndef directives, the defined operator yields a value that can be used in a preprocessor expression:

#if defined(VERSION) && defined(STATUS
...
#endif

Name

#ifdef and #ifndef

Synopsis

The #ifdef and #ifndef directives can be used to make program text directly conditional upon whether a given macro name is defined.

Syntax:

#ifdef name
#ifndef name

The #ifdef directive is "true" if name is defined, and the #ifndef directive is "true" if name is not defined. Both require a closing #endif directive.

The following two constructions are equivalent:

#ifdef VERSION
  ...
#endif

#if defined(VERSION
  ...
#endif

Name

#line

Synopsis

The compiler identifies errors it encounters during compilation by the source filename and the line number in the file. The #line directive can be used to change the filename and line numbering in the source file itself.

Syntax:

#line new_number ["filename"]

From this location in the file onward, lines are counted starting from new_number. If filename is also specified, it becomes the new filename indicated by the compiler in any error messages.

The new filename must be enclosed in quotation marks, and new_number must be an integer constant.

Example:

#line 500 "my_prg.c"

The #line directive is typically used by program generators in translating other kinds of code into a C program. In this way the C compiler's error messages can be made to refer to the appropriate line and filename in the original source code.

The current effective line number and filename are accessible through the predefined macros _ _LINE_ _ and _ _FILE_ _.

Examples:

printf( "Current source line number: %d\n", _ _LINE_ _ );
printf ( "Source file: %s\n", _ _FILE_ _ );

Name

#pragma

Synopsis

The #pragma directive is implementation-specific. It can be used to define any preprocessor directives desired for a given compiler.

Syntax:

#pragma command

Any compiler that does not recognize command simply ignores the #pragma directive.

Example:

#pragma pack(1)

The Microsoft C compiler interprets this directive as an instruction to align the members of structures on byte boundaries, so that no unnamed gaps occur. (Other pragmas supported by that compiler are pack(2) and pack(4), for word and double-word alignment.)

ANSI C99 introduces the standard pragmas CX_LIMITED_RANGE , FENV_ACCESS , and FP_CONTRACT , which are described in the upcoming section Section 1.18.

Predefined standard macros

There are eight predefined macros in C, whose names begin and end with two underline characters. They are described in Table 1-21.

Table 1-21. Predefined standard macros

ANSI C99 distinguishes between "hosted" and "free-standing" execution environments for C programs. Unlike the normal "hosted" environment, a "freestanding" environment provides only the capabilities of the standard library as declared in the header files float.h, iso646.h, limits.h, stdarg.h, stdbool.h, and stddef.h.

Standard Library

The remaining sections in this book describe the contents of the ANSI C library. The standard functions, types, and macros are grouped according to their purpose and areas of application. This arrangement makes it easy to find less well-known functions and macros. Each section also supplies the background information needed in order to make efficient use of the library's capabilities. New data types, functions, and macros introduced in ANSI C99 are indicated by an asterisk in parentheses ^(*).

Standard Header Files

All function prototypes, macros, and types in the ANSI library are contained in one or more of the following standard header files: