Why use double indirection? or Why use pointers to pointers?

Question

When should a double indirection be used in C? Can anyone explain with a example?

What I know is that a double indirection is a pointer to a pointer. Why would I ne

Answer 1

Most of the answers here are more or less related to application programming. Here is an example from embedded systems programming. For example below is an excerpt from the reference manual of NXP's Kinetis KL13 series microcontroller, this code snippet is used to run bootloader, which resides in ROM, from firmware:

" To get the address of the entry point, the user application reads the word containing the pointer to the bootloader API tree at offset 0x1C of the bootloader's vector table. The vector table is placed at the base of the bootloader's address range, which for the ROM is 0x1C00_0000. Thus, the API tree pointer is at address 0x1C00_001C.

The bootloader API tree is a structure that contains pointers to other structures, which have the function and data addresses for the bootloader. The bootloader entry point is always the first word of the API tree. "

uint32_t runBootloaderAddress;
void (*runBootloader)(void * arg);
// Read the function address from the ROM API tree.
runBootloaderAddress = **(uint32_t **)(0x1c00001c);
runBootloader = (void (*)(void * arg))runBootloaderAddress;
// Start the bootloader.
runBootloader(NULL);

Answer 2

I saw a very good example today, from this blog post, as I summarize below.

Imagine you have a structure for nodes in a linked list, which probably is

typedef struct node
{
    struct node * next;
    ....
} node;

Now you want to implement a remove_if function, which accepts a removal criterion rm as one of the arguments and traverses the linked list: if an entry satisfies the criterion (something like rm(entry)==true), its node will be removed from the list. In the end, remove_if returns the head (which may be different from the original head) of the linked list.

You may write

for (node * prev = NULL, * curr = head; curr != NULL; )
{
    node * const next = curr->next;
    if (rm(curr))
    {
        if (prev)  // the node to be removed is not the head
            prev->next = next;
        else       // remove the head
            head = next;
        free(curr);
    }
    else
        prev = curr;
    curr = next;
}

as your for loop. The message is, without double pointers, you have to maintain a prev variable to re-organize the pointers, and handle the two different cases.

But with double pointers, you can actually write

// now head is a double pointer
for (node** curr = head; *curr; )
{
    node * entry = *curr;
    if (rm(entry))
    {
        *curr = entry->next;
        free(entry);
    }
    else
        curr = &entry->next;
}

You don't need a prev now because you can directly modify what prev->next pointed to.

To make things clearer, let's follow the code a little bit. During the removal:

1. if entry == *head: it will be *head (==*curr) = *head->next -- head now points to the pointer of the new heading node. You do this by directly changing head's content to a new pointer.
2. if entry != *head: similarly, *curr is what prev->next pointed to, and now points to entry->next.

No matter in which case, you can re-organize the pointers in a unified way with double pointers.

Answer 3

I have used double pointers today while I was programming something for work, so I can answer why we had to use them (it's the first time I actually had to use double pointers). We had to deal with real time encoding of frames contained in buffers which are members of some structures. In the encoder we had to use a pointer to one of those structures. The problem was that our pointer was being changed to point to other structures from another thread. In order to use the current structure in the encoder, I had to use a double pointer, in order to point to the pointer that was being modified in another thread. It wasn't obvious at first, at least for us, that we had to take this approach. A lot of address were printed in the process :)).

You SHOULD use double pointers when you work on pointers that are changed in other places of your application. You might also find double pointers to be a must when you deal with hardware that returns and address to you.

Answer 4

1. Basic Concept -

When you declare as follows : -

1. char *ch - (called character pointer)
- ch contains the address of a single character.
- (*ch) will dereference to the value of the character..

2. char **ch -
'ch' contains the address of an Array of character pointers. (as in 1)
'*ch' contains the address of a single character. (Note that it's different from 1, due to difference in declaration).
(**ch) will dereference to the exact value of the character..

Adding more pointers expand the dimension of a datatype, from character to string, to array of strings, and so on... You can relate it to a 1d, 2d, 3d matrix..

So, the usage of pointer depends upon how you declare it.

Here is a simple code..

int main()
{
    char **p;
    p = (char **)malloc(100);
    p[0] = (char *)"Apple";      // or write *p, points to location of 'A'
    p[1] = (char *)"Banana";     // or write *(p+1), points to location of 'B'

    cout << *p << endl;          //Prints the first pointer location until it finds '\0'
    cout << **p << endl;         //Prints the exact character which is being pointed
    *p++;                        //Increments for the next string
    cout << *p;
}

2. Another Application of Double Pointers -
(this would also cover pass by reference)

Suppose you want to update a character from a function. If you try the following : -

void func(char ch)
{
    ch = 'B';
}

int main()
{
    char ptr;
    ptr = 'A';
    printf("%c", ptr);

    func(ptr);
    printf("%c\n", ptr);
}

The output will be AA. This doesn't work, as you have "Passed By Value" to the function.

The correct way to do that would be -

void func( char *ptr)        //Passed by Reference
{
    *ptr = 'B';
}

int main()
{
    char *ptr;
    ptr = (char *)malloc(sizeof(char) * 1);
    *ptr = 'A';
    printf("%c\n", *ptr);

    func(ptr);
    printf("%c\n", *ptr);
}

Now extend this requirement for updating a string instead of character.
For this, you need to receive the parameter in the function as a double pointer.

void func(char **str)
{
    strcpy(str, "Second");
}

int main()
{
    char **str;
    // printf("%d\n", sizeof(char));
    *str = (char **)malloc(sizeof(char) * 10);          //Can hold 10 character pointers
    int i = 0;
    for(i=0;i<10;i++)
    {
        str = (char *)malloc(sizeof(char) * 1);         //Each pointer can point to a memory of 1 character.
    }

    strcpy(str, "First");
    printf("%s\n", str);
    func(str);
    printf("%s\n", str);
}

In this example, method expects a double pointer as a parameter to update the value of a string.

Answer 5

The following is a very simple C++ example that shows that if you want to use a function to set a pointer to point to an object, you need a pointer to a pointer. Otherwise, the pointer will keep reverting to null.

(A C++ answer, but I believe it's the same in C.)

(Also, for reference: Google("pass by value c++") = "By default, arguments in C++ are passed by value. When an argument is passed by value, the argument's value is copied into the function's parameter.")

So we want to set the pointer b equal to the string a.

#include <iostream>
#include <string>

void Function_1(std::string* a, std::string* b) {
  b = a;
  std::cout << (b == nullptr);  // False
}

void Function_2(std::string* a, std::string** b) {
  *b = a;
  std::cout << (b == nullptr);  // False
}

int main() {
  std::string a("Hello!");
  std::string* b(nullptr);
  std::cout << (b == nullptr);  // True

  Function_1(&a, b);
  std::cout << (b == nullptr);  // True

  Function_2(&a, &b);
  std::cout << (b == nullptr);  // False
}

// Output: 10100

What happens at the line Function_1(&a, b);?

• The "value" of &main::a (an address) is copied into the parameter std::string* Function_1::a. Therefore Function_1::a is a pointer to (i.e. the memory address of) the string main::a.

• The "value" of main::b (an address in memory) is copied into the parameter std::string* Function_1::b. Therefore there are now 2 of these addresses in memory, both null pointers. At the line b = a;, the local variable Function_1::b is then changed to equal Function_1::a (= &main::a), but the variable main::b is unchanged. After the call to Function_1, main::b is still a null pointer.

What happens at the line Function_2(&a, &b);?

• The treatment of the a variable is the same: within the function, Function_2::a is the address of the string main::a.

• But the variable b is now being passed as a pointer to a pointer. The "value" of &main::b (the address of the pointer main::b) is copied into std::string** Function_2::b. Therefore within Function_2, dereferencing this as *Function_2::b will access and modify main::b . So the line *b = a; is actually setting main::b (an address) equal to Function_2::a (= address of main::a) which is what we want.

If you want to use a function to modify a thing, be it an object or an address (pointer), you have to pass in a pointer to that thing. The thing that you actually pass in cannot be modified (in the calling scope) because a local copy is made.

(An exception is if the parameter is a reference, such as std::string& a. But usually these are const. Generally, if you call f(x), if x is an object you should be able to assume that f won't modify x. But if x is a pointer, then you should assume that f might modify the object pointed to by x.)

Answer 6

Pointers to pointers also come in handy as "handles" to memory where you want to pass around a "handle" between functions to re-locatable memory. That basically means that the function can change the memory that is being pointed to by the pointer inside the handle variable, and every function or object that is using the handle will properly point to the newly relocated (or allocated) memory. Libraries like to-do this with "opaque" data-types, that is data-types were you don't have to worry about what they're doing with the memory being pointed do, you simply pass around the "handle" between the functions of the library to perform some operations on that memory ... the library functions can be allocating and de-allocating the memory under-the-hood without you having to explicitly worry about the process of memory management or where the handle is pointing.

For instance:

#include <stdlib.h>

typedef unsigned char** handle_type;

//some data_structure that the library functions would work with
typedef struct 
{
    int data_a;
    int data_b;
    int data_c;
} LIB_OBJECT;

handle_type lib_create_handle()
{
    //initialize the handle with some memory that points to and array of 10 LIB_OBJECTs
    handle_type handle = malloc(sizeof(handle_type));
    *handle = malloc(sizeof(LIB_OBJECT) * 10);

    return handle;
}

void lib_func_a(handle_type handle) { /*does something with array of LIB_OBJECTs*/ }

void lib_func_b(handle_type handle)
{
    //does something that takes input LIB_OBJECTs and makes more of them, so has to
    //reallocate memory for the new objects that will be created

    //first re-allocate the memory somewhere else with more slots, but don't destroy the
    //currently allocated slots
    *handle = realloc(*handle, sizeof(LIB_OBJECT) * 20);

    //...do some operation on the new memory and return
}

void lib_func_c(handle_type handle) { /*does something else to array of LIB_OBJECTs*/ }

void lib_free_handle(handle_type handle) 
{
    free(*handle);
    free(handle); 
}


int main()
{
    //create a "handle" to some memory that the library functions can use
    handle_type my_handle = lib_create_handle();

    //do something with that memory
    lib_func_a(my_handle);

    //do something else with the handle that will make it point somewhere else
    //but that's invisible to us from the standpoint of the calling the function and
    //working with the handle
    lib_func_b(my_handle); 

    //do something with new memory chunk, but you don't have to think about the fact
    //that the memory has moved under the hood ... it's still pointed to by the "handle"
    lib_func_c(my_handle);

    //deallocate the handle
    lib_free_handle(my_handle);

    return 0;
}

Hope this helps,

Jason