struct (C programming language)

In other languages

Like its C counterpart, the struct data type in C# (Structure in Visual Basic .NET) is similar to a class. The biggest difference between a struct and a class in these languages is that when a struct is passed as an argument to a function, any modifications to the struct in that function will not be reflected in the original variable (unless pass-by-reference is used).^[2]

This distinction differs from C++, where classes or structs can be allocated either on the stack (similar to C#) or on the heap, with an explicit pointer. In C++, the only difference between a struct and a class is that the members and base classes of a struct are public by default. (A class defined with the class keyword has private members and base classes by default.)

Declaration

The general syntax for a struct declaration in C is:

struct tag_name {
   type member1;
   type member2;
   /* declare as many members as desired, but the entire structure size
   must be known to the compiler. */
};

Here tag_name is optional in some contexts.

Such a struct declaration may also appear in the context of a typedef declaration of a type alias or the declaration or definition of a variable:

typedef struct tag_name {
   type member1;
   type member2;
} struct_alias;

Often, such entities are better declared separately, as in:

typedef struct tag_name struct_alias;
 
// These two statements now have the same meaning:
// struct tag_name struct_instance;
// struct_alias struct_instance;

For example:

struct account {
   int account_number;
   char *first_name;
   char *last_name;
   float balance;
};

defines a type, referred to as struct account. To create a new variable of this type, we can write

struct account s;

which has an integer component, accessed by s.account_number, and a floating-point component, accessed by s.balance, as well as the first_name and last_name components. The structure s contains all four values, and all four fields may be changed independently.

A pointer to an instance of the "account" structure will point to the memory address of the first variable, "account_number". The total storage required for a struct object is the sum of the storage requirements of all the fields, plus any internal padding.

The primary use of a struct is for the construction of complex data types, but in practice they are sometimes used to circumvent standard C conventions to create a kind of primitive subtyping. For example, common Internet protocols rely on the fact that C compilers insert padding between struct fields in predictable ways; thus the code

struct ifoo_version_42 {
   long x, y, z;
   char *name;
   long a, b, c;
};
struct ifoo_old_stub {
   long x, y;
};
void operate_on_ifoo(struct ifoo_version_42 *);
struct ifoo_old_stub s;
. . .
operate_on_ifoo(&s);

is often assumed to work as expected, if the operate_on_ifoo function only accesses fields x and y of its argument because a C compiler will map the struct elements to memory space exactly as it is written in the source code, thus the x and y will point to exactly the same memory space in both structs.

Struct initialization

There are three ways to initialize a structure. For the struct type

/* Forward declare a type "point" to be a struct. */
typedef struct point point;
/* Declare the struct with integer members x, y */
struct point {
   int    x;
   int    y;
};

C89-style initializers are used when contiguous members may be given.^[3]

/* Define a variable p of type point, and initialize its first two members in place */
point p = {1,2};

For non contiguous or out of order members list, designated initializer style^[4] may be used

/* Define a variable p of type point, and set members using designated  initializers*/
point p = {.y = 2, .x = 1};

If an initializer is given or if the object is statically allocated, omitted elements are initialized to 0.^[5]

A third way of initializing a structure is to copy the value of an existing object of the same type

/* Define a variable q of type point, and set members to the same values as those of p */
point q = p;

Assignment

The following assignment of a struct to another struct will copy as one might expect. Depending on the contents, a compiler might use memcpy() in order to perform this operation.

#include <stdio.h>
 
/* Define a type point to be a struct with integer members x, y */
typedef struct {
   int    x;
   int    y;
} point;
 
int main(void) {
 
/* Define a variable p of type point, and initialize all its members inline! */
    point p = {1,3};
 
/* Define a variable q of type point. Members are uninitialized. */
    point q;
 
/* Assign the value of p to q, copies the member values from p into q. */
    q = p;
 
/* Change the member x of q to have the value of 3 */
    q.x = 3;
 
/* Demonstrate we have a copy and that they are now different. */
    if (p.x != q.x) printf("The members are not equal! %d != %d", p.x, q.x);
 
    return 0;
}

Pointers to struct

Pointers can be used to refer to a struct by its address. This is particularly useful for passing structs to a function by reference or to refer to another instance of the struct type as a field. The pointer can be dereferenced just like any other pointer in C — using the * operator. There is also a -> operator in C which dereferences the pointer to struct (left operand) and then accesses the value of a member of the struct (right operand).

struct point {
   int x;
   int y;
};
struct point my_point = { 3, 7 };
struct point *p = &my_point;  /* To declare and define p as a pointer of type struct point,
                                 and initialize it with the address of my_point. */
 
(*p).x = 8;                   /* To access the first member of the struct */
p->x = 8;                     /* Another way to access the first member of the struct */

C does not allow recursive declaration of struct; a struct can not contain a field that has the type of the struct itself. But pointers can be used to refer to an instance of it:

typedef struct list_element list_element;
struct list_element {
   point p;
   list_element * next;
};
list_element el = { .p = { .x = 3, .y =7 }, };
list_element le = { .p = { .x = 4, .y =5 }, .next = &el };

Here the instance el would contain a point with coordinates 3 and 7. Its next pointer would be a null pointer since the initializer for that field is omitted. The instance le in turn would have its own point and its next pointer would refer to el.

typedef

Main article: typedef

Typedefs can be used as shortcuts, for example:

typedef struct {
   int    account_number;
   char   *first_name;
   char   *last_name;
   float  balance;
} account;

Different users have differing preferences; proponents usually claim:

shorter to write
can simplify more complex type definitions
can be used to forward declare a struct type

As an example, consider a type that defines a pointer to a function that accepts pointers to struct types and returns a pointer to struct:

Without typedef:

struct point {
   int    x;
   int    y;
};
struct point *(*point_compare_func) (struct point *a, struct point *b);

With typedef:

typedef struct point point_type;
struct point {
   int    x;
   int    y;
};
point_type *(*point_compare_func) (point_type *a, point_type *b);

A common naming convention for such a typedef is to append a "_t" (here point_t) to the struct tag name, but such names are reserved by POSIX so such a practice should be avoided. A much easier convention is to use just the same identifier for the tag name and the type name:

typedef struct point point;
struct point {
   int    x;
   int    y;
};
point *(*point_compare_func) (point *a, point *b);

Without typedef a function that takes function pointer the following code would have to be used. Although valid, it becomes increasingly hard to read.

/* Using the struct point type from before */
 
/* Define a function that returns a pointer to the biggest point,
   using a function to do the comparison. */
struct point *
biggest_point (size_t size, struct point *points,
               struct point *(*point_compare) (struct point *a, struct point *b))
{
    int i;
    struct point *biggest = NULL;
 
    for (i=0; i < size; i++) {
        biggest = point_compare(biggest, points + i);
    }
    return biggest;
}

Here a second typedef for a function pointer type can be useful

typedef point *(*point_compare_func_type) (point *a, point *b);

Now with the two typedefs being used the complexity of the function signature is drastically reduced.

/* Using the struct point type from before and the typedef for the function pointer */
 
/* Define a function that returns a pointer to the biggest point,
   using a function to do the comparison. */
point *
biggest_point (size_t size, point * points, point_compare_func_type point_compare)
{
    int i;
    point * biggest = NULL;
 
    for (i=0; i < size; i++) {
        biggest = point_compare(biggest, points + i);
    }
    return biggest;
}

However, there are a handful of disadvantages in using them:

They pollute the main namespace (see below), however this is easily overcome with prefixing a library name to the type name.
Harder to figure out the aliased type (having to scan/grep through code), though most IDEs provide this lookup automatically.
Typedefs do not really "hide" anything in a struct or union — members are still accessible (account.balance). To really hide struct members, one needs to use 'incompletely-declared' structs.

/* Example for namespace clash */
 
typedef struct account { float balance; } account;
struct account account; /* possible */
account account; /* error */

References

↑ C struct memory layout? - Stack Overflow
↑ Parameter passing in C#
↑ Kelley, Al; Pohl, Ira (2004). A Book On C: Programming in C (Fourth ed.). p. 418. ISBN 0-201-18399-4.
↑ "IBM Linux compilers. Initialization of structures and unions".
↑ "The New C Standard, §6.7.8 Initialization".