Is an array name a pointer?

Is an array's name a pointer in C? If 开发者_如何学运维not, what is the difference between an array's name and a pointer variable?

---

An array is an array and a pointer is a pointer, but in most cases array names are converted to pointers. A term often used is that they decay to pointers.

Here is an array:

int a[7];

a contains space for seven integers, and you can put a value in one of them with an assignment, like this:

a[3] = 9;

Here is a pointer:

int *p;

p doesn't contain any spaces for integers, but it can point to a space for an integer. We can, for example, set it to point to one of the places in the array a, such as the first one:

p = &a[0];

What can be confusing is that you can also write this:

p = a;

This does not copy the contents of the array a into the pointer p (whatever that would mean). Instead, the array name a is converted to a pointer to its first element. So that assignment does the same as the previous one.

Now you can use p in a similar way to an array:

p[3] = 17;

The reason that this works is that the array dereferencing operator in C, [ ], is defined in terms of pointers. x[y] means: start with the pointer x, step y elements forward after what the pointer points to, and then take whatever is there. Using pointer arithmetic syntax, x[y] can also be written as *(x+y).

For this to work with a normal array, such as our a, the name a in a[3] must first be converted to a pointer (to the first element in a). Then we step 3 elements forward, and take whatever is there. In other words: take the element at position 3 in the array. (Which is the fourth element in the array, since the first one is numbered 0.)

So, in summary, array names in a C program are (in most cases) converted to pointers. One exception is when we use the sizeof operator on an array. If a was converted to a pointer in this context, sizeof a would give the size of a pointer and not of the actual array, which would be rather useless, so in that case a means the array itself.

---

When an array is used as a value, its name represents the address of the first element.
When an array is not used as a value its name represents the whole array.

int arr[7];

/* arr used as value */
foo(arr);
int x = *(arr + 1); /* same as arr[1] */

/* arr not used as value */
size_t bytes = sizeof arr;
void *q = &arr; /* void pointers are compatible with pointers to any object */

---

If an expression of array type (such as the array name) appears in a larger expression and it isn't the operand of either the & or sizeof operators, then the type of the array expression is converted from "N-element array of T" to "pointer to T", and the value of the expression is the address of the first element in the array.

In short, the array name is not a pointer, but in most contexts it is treated as though it were a pointer.

Edit

Answering the question in the comment:

If I use sizeof, do i count the size of only the elements of the array? Then the array “head” also takes up space with the information about length and a pointer (and this means that it takes more space, than a normal pointer would)?

When you create an array, the only space that's allocated is the space for the elements themselves; no storage is materialized for a separate pointer or any metadata. Given

char a[10];

what you get in memory is

   +---+
a: |   | a[0]
   +---+ 
   |   | a[1]
   +---+
   |   | a[2]
   +---+
    ...
   +---+
   |   | a[9]
   +---+

The expression a refers to the entire array, but there's no object a separate from the array elements themselves. Thus, sizeof a gives you the size (in bytes) of the entire array. The expression &a gives you the address of the array, which is the same as the address of the first element. The difference between &a and &a[0] is the type of the result¹ - char (*)[10] in the first case and char * in the second.

Where things get weird is when you want to access individual elements - the expression a[i] is defined as the result of *(a + i) - given an address value a, offset i elements (not bytes) from that address and dereference the result.

The problem is that a isn't a pointer or an address - it's the entire array object. Thus, the rule in C that whenever the compiler sees an expression of array type (such as a, which has type char [10]) and that expression isn't the operand of the sizeof or unary & operators, the type of that expression is converted ("decays") to a pointer type (char *), and the value of the expression is the address of the first element of the array. Therefore, the expression a has the same type and value as the expression &a[0] (and by extension, the expression *a has the same type and value as the expression a[0]).

C was derived from an earlier language called B, and in B a was a separate pointer object from the array elements a[0], a[1], etc. Ritchie wanted to keep B's array semantics, but he didn't want to mess with storing the separate pointer object. So he got rid of it. Instead, the compiler will convert array expressions to pointer expressions during translation as necessary.

Remember that I said arrays don't store any metadata about their size. As soon as that array expression "decays" to a pointer, all you have is a pointer to a single element. That element may be the first of a sequence of elements, or it may be a single object. There's no way to know based on the pointer itself.

When you pass an array expression to a function, all the function receives is a pointer to the first element - it has no idea how big the array is (this is why the gets function was such a menace and was eventually removed from the library). For the function to know how many elements the array has, you must either use a sentinel value (such as the 0 terminator in C strings) or you must pass the number of elements as a separate parameter.

---

1. Which *may* affect how the address value is interpreted - depends on the machine.

---

An array declared like this

int a[10];

allocates memory for 10 ints. You can't modify a but you can do pointer arithmetic with a.

A pointer like this allocates memory for just the pointer p:

int *p;

It doesn't allocate any ints. You can modify it:

p = a;

and use array subscripts as you can with a:

p[2] = 5;
a[2] = 5;    // same
*(p+2) = 5;  // same effect
*(a+2) = 5;  // same effect

---

The array name by itself yields a memory location, so you can treat the array name like a pointer:

int a[7];

a[0] = 1976;
a[1] = 1984;

printf("memory location of a: %p", a);

printf("value at memory location %p is %d", a, *a);

And other nifty stuff you can do to pointer (e.g. adding/substracting an offset), you can also do to an array:

printf("value at memory location %p is %d", a + 1, *(a + 1));

Language-wise, if C didn't expose the array as just some sort of "pointer"(pedantically it's just a memory location. It cannot point to arbitrary location in memory, nor can be controlled by the programmer). We always need to code this:

printf("value at memory location %p is %d", &a[1], a[1]);

---

I think this example sheds some light on the issue:

#include <stdio.h>
int main()
{
        int a[3] = {9, 10, 11};
        int **b = &a;

        printf("a == &a: %d\n", a == b);
        return 0;
}

It compiles fine (with 2 warnings) in gcc 4.9.2, and prints the following:

a == &a: 1

oops :-)

So, the conclusion is no, the array is not a pointer, it is not stored in memory (not even read-only one) as a pointer, even though it looks like it is, since you can obtain its address with the & operator. But - oops - that operator does not work :-)), either way, you've been warned:

p.c: In function ‘main’:
pp.c:6:12: warning: initialization from incompatible pointer type
  int **b = &a;
            ^
p.c:8:28: warning: comparison of distinct pointer types lacks a cast
  printf("a == &a: %d\n", a == b);

C++ refuses any such attempts with errors in compile-time.

Edit:

This is what I meant to demonstrate:

#include <stdio.h>
int main()
{
    int a[3] = {9, 10, 11};
    void *c = a;

    void *b = &a;
    void *d = &c;

    printf("a == &a: %d\n", a == b);
    printf("c == &c: %d\n", c == d);
    return 0;
}

Even though c and a "point" to the same memory, you can obtain address of the c pointer, but you cannot obtain the address of the a pointer.

---

The following example provides a concrete difference between an array name and a pointer. Let say that you want to represent a 1D line with some given maximum dimension, you could do it either with an array or a pointer:

typedef struct {
   int length;
   int line_as_array[1000];
   int* line_as_pointer;
} Line;

Now let's look at the behavior of the following code:

void do_something_with_line(Line line) {
   line.line_as_pointer[0] = 0;
   line.line_as_array[0] = 0;
}

void main() {
   Line my_line;
   my_line.length = 20;
   my_line.line_as_pointer = (int*) calloc(my_line.length, sizeof(int));

   my_line.line_as_pointer[0] = 10;
   my_line.line_as_array[0] = 10;

   do_something_with_line(my_line);

   printf("%d %d\n", my_line.line_as_pointer[0], my_line.line_as_array[0]);
};

This code will output:

0 10

That is because in the function call to do_something_with_line the object was copied so:

1. The pointer line_as_pointer still contains the same address it was pointing to
2. The array line_as_array was copied to a new address which does not outlive the scope of the function

So while arrays are not given by values when you directly input them to functions, when you encapsulate them in structs they are given by value (i.e. copied) which outlines here a major difference in behavior compared to the implementation using pointers.

---

The array name behaves like a pointer and points to the first element of the array. Example:

int a[]={1,2,3};
printf("%p\n",a);     //result is similar to 0x7fff6fe40bc0
printf("%p\n",&a[0]); //result is similar to 0x7fff6fe40bc0

Both the print statements will give exactly same output for a machine. In my system it gave:

0x7fff6fe40bc0