C Programming - Pointers

A void pointer is a C convention for "a raw address." The compiler has no idea what type of object a void pointer "really points to." If you write

int *ip;

ip points to an int. If you write

void *p;

p doesn't point to a void!

In C and C++, any time you need a void pointer, you can use another pointer type. For example, if you have a char*, you can pass it to a function that expects a void*. You don't even need to cast it. In C (but not in C++), you can use a void* any time you need any kind of pointer, without casting. (In C++, you need to cast it.)

A void pointer is used for working with raw memory or for passing a pointer to an unspecified type.

Some C code operates on raw memory. When C was first invented, character pointers (char *) were used for that. Then people started getting confused about when a character pointer was a string, when it was a character array, and when it was raw memory.

For example, strcpy() is used to copy data from one string to another, and strncpy() is used to copy at most a certain length string to another:

char *strcpy( char *str1, const char *str2 );

char *strncpy( char *str1, const char *str2, size_t n );

memcpy() is used to move data from one location to another:

void *memcpy( void *addr1, void *addr2, size_t n );

void pointers are used to mean that this is raw memory being copied. NUL characters (zero bytes) aren't significant, and just about anything can be copied. Consider the following code:

#include "thingie.h"    /* defines struct thingie */
struct thingie  *p_src, *p_dest;
/* ... */
memcpy( p_dest, p_src, sizeof( struct thingie) * numThingies );

This program is manipulating some sort of object stored in a struct thingie. p1 and p2 point to arrays, or parts of arrays, of struct thingies. The program wants to copy numThingies of these, starting at the one pointed to by p_src, to the part of the array beginning at the element pointed to by p_dest. memcpy() treats p_src and p_dest as pointers to raw memory; sizeof( struct thingie) * numThingies is the number of bytes to be copied.

If you have two pointers into the same array, you can subtract them. The answer is the number of elements between the two elements.

Consider the street address analogy presented in the introduction of this chapter. Say that I live at 118 Fifth Avenue and that my neighbor lives at 124 Fifth Avenue. The "size of a house" is two (on my side of the street, sequential even numbers are used), so my neighbor is (124-118)/2 (or 3) houses up from me. (There are two houses between us, 120 and 122; my neighbor is the third.) You might do this subtraction if you're going back and forth between indices and pointers.

You might also do it if you're doing a binary search. If p points to an element that's before what you're looking for, and q points to an element that's after it, then (q-p)/2+p points to an element between p and q. If that element is before what you want, look between it and q. If it's after what you want, look between p and it.

(If it's what you're looking for, stop looking.)

You can't subtract arbitrary pointers and get meaningful answers. Someone might live at 110 Main Street, but I can't subtract 110 Main from 118 Fifth (and divide by 2) and say that he or she is four houses away!

If each block starts a new hundred, I can't even subtract 120 Fifth Avenue from 204 Fifth Avenue. They're on the same street, but in different blocks of houses (different arrays).

C won't stop you from subtracting pointers inappropriately. It won't cut you any slack, though, if you use the meaningless answer in a way that might get you into trouble.

When you subtract pointers, you get a value of some integer type. The ANSI C standard defines a typedef, ptrdiff_t, for this type. (It's in <stddef.h>.) Different compilers might use different types (int or long or whatever), but they all define ptrdiff_t appropriately.

Below is a simple program that demonstrates this point. The program has an array of structures, each 16 bytes long. The difference between array[0] and array[8] is 8 when you subtract struct stuff pointers, but 128 (hex 0x80) when you cast the pointers to raw addresses and then subtract.

If you subtract 8 from a pointer to array[8], you don't get something 8 bytes earlier; you get something 8 elements earlier.

#include <stdio.h>
#include <stddef.h>
struct stuff {
        char    name[16];
        /* other stuff could go here, too */
};
struct stuff array[] = {
        { "The" },
        { "quick" },
        { "brown" },
        { "fox" },
        { "jumped" },
        { "over" },
        { "the" },
        { "lazy" },
        { "dog." },
        { "" }
};
int main()
{
        struct stuff    *p0 = & array[0];
        struct stuff    *p8 = & array[8];
        ptrdiff_t       diff = p8 - p0;
        ptrdiff_t       addr_diff = (char*) p8 - (char*) p0;
        /* cast the struct stuff pointers to void* */
        printf("& array[0] = p0 = %P\n", (void*) p0);
        printf("& array[8] = p8 = %P\n", (void*) p8);
        /* cast the ptrdiff_t's to long's
        (which we know printf() can handle) */
        printf("The difference of pointers is %ld\n",
          (long) diff);
        printf("The difference of addresses is %ld\n",
          (long) addr_diff);
        printf("p8 - 8 = %P\n", (void*) (p8 - 8));
        
        printf("p0 + 8 = %P (same as p8)\n", (void*) (p0 + 8));
        return 0;  
}

NULL is defined as either 0 or (void*)0. These values are almost identical; either a literal zero or a void pointer is converted automatically to any kind of pointer, as necessary, whenever a pointer is needed (although the compiler can't always tell when a pointer is needed).