UP | HOME

8. Pointers

Table of Contents


Pointers

Pointers represent one of the more powerful features of the C language, but also one of the most feared. Some of this fear and apprehension stems from the ridiculously obtuse ways that pointers may be used in C. Often tutorials and courses on C approach teaching pointers by asking the student to decipher bizarre looking pointer syntax combinations that truthfully are rarely used in practice. You may have seen bizarre looking pointer syntax, (and you may see it again), but typically, pointers do not have to be used in a horribly complicated way. Typically pointers are pretty straightforward.

The bottom line is, pointers allow you to work with dynamically allocated memory blocks. You will use pointers to deal with variables that have been allocated on the heap. (You can use pointers to deal with stack variables, but in most cases this just isn't necessary).

The basic idea is, a pointer is a special data type in C, that contains an address to a location in memory. Think of a pointer as an arrow that points to the location of a block of memory that in turn contains actual data of interest. It's not unlike the concept of a phone number, or a house address.

The purpose of pointers is to allow you to manually, directly access a block of memory. Pointers are used a lot for strings and structs. It's not difficult to imagine that passing the address of a large block of memory (such as a struct that contains many things) to a function, is more efficient than making a copy of it and passing in that copy, only to delete that copy when your function is done with it. This is known as passing by reference versus passing by value. There is a code example below that illustrates this more clearly.

Declaring a pointer

Here is how to declare a pointer:

#include <stdio.h>

int main (int argc, char *argv[])
{
  int age = 30;
  int *p;
  p = &age;
  printf("age=%d\n", age);
  printf("p=%p\n", p);
  printf("*p=%d\n", *p);
  printf("sizeof(p)=%ld\n", sizeof(p));
  *p = 40;
  printf("*p=%d\n", *p);
  printf("age=%d\n", age);
  return 0;
}
age=30
p=0x7fff197ceb1c
*p=30
sizeof(p)=8
*p=40
age=40

On line 5 we declare an int variable age and initialize it to 30. On line 6 we declare a variable p whose type is a pointer to an int. The star * syntax is how to declare the type as a pointer to a particular other type. Usually you need to declare pointers as pointing to a particular type, but the exception is a so-called void pointer, which is a pointer to an unknown type. Don't worry about void pointers for now.

So we have a pointer variable p that is a pointer to an int. So far, it doesn't point anywhere… we have only declared that we want a space in memory to hold a pointer to an int. On line 7 is where we assign a value to our pointer p. We assign to p the address (the location) of the other variable age. So now at this point in the program, we have the following (which we can see in the output):

  • the value of age is the integer 30
  • the value of p is the address of the age variable
  • the integer that p points to has a value of 30

On line 2 of the output we see that the value of the p pointer is a long strange looking string of letters and numbers. This is in fact a hexidecimal (base 16) memory address.

Dereferencing a Pointer

On line 10, we see how to access the value that a pointer points to, by using the star * syntax. So p is the address of an int, whereas *p is the value of the int that p points to. Accessing the value that a pointer points to is called dereferencing a pointer.

Finally, on line 11 (and on the last line of the output) we can see that our pointer p is 8 bytes. Remember, there are 8 bits in a byte, so a 4-byte pointer can hold up to 32 bits, or \(2^{32}\) distinct values, i.e. \(2^{32}\) distinct addresses in memory. So 32-bit pointers allow us to access up to 4 gigabytes (4 GB) of memory. If you want to use more than 4 GB of memory you will need bigger pointers! On 64-bit systems, pointers are 8 bytes, which allows for \(2^{64}\) distinct addresses in memory. It will be a long time before your computer has that much RAM (16 exabytes in principle, which is 1 billion gigabytes or 1 million terabytes).

Take note of what happens in the code example above on line 12 of the code listing, and the values output on lines 5 and 6 of the output. On line 12 of the code listing, we use pointer dereferencing to set the value of the variable pointed to by p to 40. Since p points to the age variable, we are setting the value of the address in memory corresponding to the age variable to 40.

Pointers and Arrays

When you declare an array using an expression like int vec[5];, what is really happening behind the scenes, is that a block of memory is being allocated (on the stack in this case) large enough to hold 5 integers, and the vec variable is a pointer that points to the first element in the array. When you index into the array with an expression like printf("vec[2]=%d\n", vec[2]); what is really happening is that C is using pointer arithmetic to step into the array the appropriate number of times, and reading the value in the memory location it ends up in. So if you ask for the 3rd element of the vec array using vec[2] then C is first looking at the location pointed to by vec (the first element of the array), and stepping two integers forward, and then reading the value it finds there (vec[2]).

The explicit way of doing this would be to use malloc() or calloc() to allocate the array (in this case on the heap) and then use pointer arithmetic to read off the 3rd value:

#include <stdio.h>
#include <stdlib.h>

int main(void) {
  int *vec;
  vec = malloc(sizeof(int) * 3);
  vec[0] = 1;
  vec[1] = 2;
  vec[2] = 3;
  printf("vec[2]=%d\n", *(vec+2));
  free(vec);
  return 0;
}

The expression *(vec+2) essentially says, go to the location that the pointer vec points to, take two steps (each step the size of an int) and the read off the value you find there. How does C know the size of each step to take? Remember when you declare a pointer, you have to declare the type that it points to. So when we declared our vec pointer, we declared it as a pointer to int. This means when you use pointer arithmetic, C knows the size of each step.

The *(vec+2) notation is really just as a demonstration of what is happening behind the scenes, there is rarely a need to use such an explicit (and so difficult to read) expression to index into arrays. The more common way of indexing into arrays (and the more readable way) is to use the vec[2] notation:

#include <stdio.h>
#include <stdlib.h>

int main(void) {
  int *vec;
  vec = malloc(sizeof(int) * 3);
  vec[0] = 1;
  vec[1] = 2;
  vec[2] = 3;
  printf("vec[2]=%d\n", vec[2]);
  free(vec);
  return 0;
}

Pointer to a struct

Pointers can also be used to point to a struct. Here is how this would be done:

#include <stdio.h>
#include <stdlib.h>

typedef struct {
  int year;
  int month;
  int day;
} date;

int main(void) {

  date *today;
  today = malloc(sizeof(date));

  // the explicit way of accessing fields of our struct
  (*today).day = 15;
  (*today).month = 6;
  (*today).year = 2012;

  // the more readable shorthand way of doing it
  today->day = 15;
  today->month = 6;
  today->year = 2012;

  printf("the date is %d/%d/%d\n", today->day, today->month, today->year);

  free(today);

  return 0;
}
the date is 15/6/2012

Let's go through this step by step. On lines 4-8 we define a struct that contains three int values: year, month and day. We use typedef to name our new struct type date.

On line 12 we declare a new variable today to be a pointer to date. On line 13 we use malloc() to allocate a block of memory (on the heap) to store one date struct.

On lines 16-18 I show how to access fields of our date struct, using explicit pointer syntax. So for example the expression (*today).day means, dereference the today pointer and then access the day field of the thing you find there (which will be a date struct).

On lines 21-23 I show you the more common (and more readable) shorthand for using pointers with structs.

Just as a reminder: here is how one would do this on the stack instead of the heap:

#include <stdio.h>

typedef struct {
  int year;
  int month;
  int day;
} date;

int main(void) {

  date today;

  today.day = 15;
  today.month = 6;
  today.year = 2012;

  printf("the date is %d/%d/%d\n", today.day, today.month, today.year);

  return 0;
}

Gone is all of the pointer stuff, at least on the surface. Under the hood, C is still using pointers to accomplish this.

Pointers to Functions

One of the handy things you can do in C, is to use a pointer to point to a function. Then you can pass this function pointer to other functions as an argument, you can store it in a struct, etc. Here is a small example:

#include <stdio.h>

int add( int a, int b ) {
  return a + b;
}

int subtract( int a, int b ) {
  return a - b;
}

int multiply( int a, int b ) {
  return a * b;
}

void doMath( int (*fn)(int a, int b), int a, int b ) {
  int result = fn(a, b);
  printf("result = %d\n", result);
}

int main(void) {

  int a = 2;
  int b = 3;

  doMath(add, a, b);
  doMath(subtract, a, b);
  doMath(multiply, a, b);

  return 0;
}

Let's go through this to understand what's happening. On lines 3-5, 7-9 and 11-13, we define functions add, subtract and multiply. These functions return an int and take two int values as input arguments.

On lines 15-18 we define a function doMath which returns nothing (hence void) and which takes three input arguments. The first input argument is:

int (*fn)(int a, int b)

This first argument is a pointer to a function. It's actually more specific than that. It's a pointer to a specific kind of function: a function that returns an int, and that takes two int values as inputs. Let's unpack this. The (*fn) says this is a pointer to a function, and we shall refer to that function as fn. The preceding int says, it's a function that returns an int. The subsequent (int a, int b) says it's a function that takes two int arguments as inputs.

On lines 25-27, we call our doMath() function, each time passing it the three input arguments that it requires. First, a pointer to a function. Here we simply pass the name of one of the functions we defined above: add(), subtract() or multiply(). We are permitted to pass these functions to doMath() because they all satisfy the requirements of the first input argument of doMath(): they all return an int, and they all take two int values as inputs.

Function Arguments: Passing By Value vs Passing By Reference

Typically when you think about passing arguments to functions, you think about passing the function the value of some variable. A common idiom in C however is to pass function arguments by reference, using pointers. This is the case in particular with large data structures like arrays and structs, for which it would be inefficient to make a copy of the whole thing, and passing that copy to a function. Instead, in passing by reference, you simply pass a pointer to the data, to the function.

Here is some code illustrating passing by value, first:

#include <stdio.h>

void myFun(int x) {
  x = x * 2;
}

int main(void) {
  int y = 50;
  printf("y=%d\n", y);
  myFun(y);
  printf("y=%d\n", y);
  return 0;
}
y=50
y=50

In the above code example, on line 8, we assign the value 50 to the variable y. Then on line 10 we call the function myFun() which takes one int argument and we pass it our variable y. This is passing by value since we are handing over to myFun() the value of y (50). Within myFun() we multiply the argument passed to it by 2 and exit. In main() when we print the value of y, after the function call to myFun(), it is still 50 (not 100).

#include <stdio.h>

void myFun(int *x) {
  *x = *x * 2;
}

int main(void) {
  int y = 50;
  printf("y=%d\n", y);
  myFun(&y);
  printf("y=%d\n", y);
  return 0;
}
y=50
y=100

Here, we rewrite myFun() so that it take an input argument that is not an int, but rather a pointer to an int (hence the star * notation). Now in our main() function, on line 10, we pass to myFun() not the value of y as in the previous code example, but rather the address of y, using the & notation. Now when myFun() is called, it uses pointer dereferencing to multiply the value pointed to by its argument x, by 2. Of course x is simply the address of y, which we passed to myFun(), and so the value pointed to by x is the value that we assigned to y, which is 50. So myFun() multiplies that value by 2 and assigns it using pointer dereferencing to *x, which is the value associated with y.

Make sure you understand these two code examples above, and why they do different things. If you understand this, then you basically understand pointers.

Dynamically Allocated Memory

Languages like MATLAB, Python, etc, allow you to work with data structures like arrays, lists, etc, that you can dynamically resize. That is to say, you can make them longer, shorter, etc, even after they are created. In C this is not so easy.

Once you have allocated a variable such as an array on the stack, it is fixed in its size. You cannot make it longer or shorter. In contrast, if you use malloc() or calloc() to allocate an array on the heap, you can use realloc() to resize it at some later time. In order to use these functions you will need to #include <stdlib.h> at the top of your C file.

The built-in functions malloc(), calloc(), realloc() memcpy() and free() are what you will use to manage dynamically allocated data structures on the heap, "by hand". The life cycle of a heap variable involves three stages:

  1. allocating the heap variable using malloc() or calloc()
  2. (optionally) resizing the heap variable using realloc()
  3. releasing the memory from the heap using free()

Allocating memory using malloc() or calloc()

These functions are used to allocate memory at runtime. The malloc() function takes as input the size of the memory block to be allocated. The calloc() function is like malloc() except that it also initializes all elements to zero. The calloc() function takes two input arguments, the number of elements and the size of each element.

Here's an example of using malloc() to allocate memory to hold an array of 10 structs:

#include <stdio.h>
#include <stdlib.h>

typedef struct {
  int year;
  int month;
  int day;
} date;

int main(void) {

  date *mylist = malloc(sizeof(date) * 10);

  mylist[0].year = 2012;
  mylist[0].month = 1;
  mylist[0].day = 15;

  int i;
  for (i=1; i<10; i++) {
    mylist[i].year = 2012-i;
    mylist[i].month = 1 + i;
    mylist[i].day = 15 + i;
  }

  for (i=0; i<10; i++) {
    printf("mylist[%d] = %d/%d/%d\n", i, mylist[i].day, mylist[i].month, mylist[i].year);
  }

  free(mylist);
  return 0;
}
mylist[0] = 15/1/2012
mylist[1] = 16/2/2011
mylist[2] = 17/3/2010
mylist[3] = 18/4/2009
mylist[4] = 19/5/2008
mylist[5] = 20/6/2007
mylist[6] = 21/7/2006
mylist[7] = 22/8/2005
mylist[8] = 23/9/2004
mylist[9] = 24/10/2003

Resizing a variable using realloc()

Let's say you use calloc() to allocate an array of 3 floating-point values, and you later in the program want to increase the size of the array to hold 5 values. Here's how you could do it using realloc():

#include <stdio.h>
#include <stdlib.h>

void showVec(double vec[], int n) {
  int i;
  for (i=0; i<n; i++) {
    printf("vec[%d]=%.3f\n", i, vec[i]);
  }
  printf("\n");
}

int main(void) {

  double *vec = calloc(3, sizeof(double));

  vec[1] = 3.14;
  showVec(vec, 3);

  vec = realloc(vec, sizeof(double)*5);
  showVec(vec, 5);

  vec[3] = 7.77;
  showVec(vec, 5);

  free(vec);
  return 0;
}
vec[0]=0.000
vec[1]=3.140
vec[2]=0.000

vec[0]=0.000
vec[1]=3.140
vec[2]=0.000
vec[3]=0.000
vec[4]=0.000

vec[0]=0.000
vec[1]=3.140
vec[2]=0.000
vec[3]=7.770
vec[4]=0.000

Freeing a memory block using free()

You should always use free() to deallocate memory that has been allocated with malloc() or calloc(), as soon as you don't need it any more. Any memory allocated with malloc() or calloc() is reserved, in other words, it can't be used (for good reason) until it is deallocated with free(). If you fail to deallocate memory then you will have what's called a memory leak. If your program uses a lot of heap memory, that is not deallocated, and runs for a long time, then you might find that your computer (and your program) slows down, or suddenly freezes, or crashes, because there is no more memory to be allocated.

The rule is, anytime you use malloc() or calloc(), you must also use free().

Links

Exercises

  • 1 Refactor the code from your matrix program (from here) so that the size of a matrix is not fixed to a maximum number of elements. Instead use dynamic memory allocation.

    Here are some hints:

#include <stdio.h>
#include <stdlib.h>

typedef struct {
  double *data;
  int nrows;
  int ncols;
} Matrix;

void printmat(Matrix *M);
void matrixmult(Matrix *A, Matrix *B, Matrix *C);
Matrix *createMatrix(int nrows, int ncols);
void destroyMatrix(Matrix *M);

int main(int argc, char *argv[])
{
  Matrix *A = createMatrix(3, 2);
  A->data[0] = 1.2;
  A->data[1] = 2.3;
  A->data[2] = 3.4;
  A->data[3] = 4.5;
  A->data[4] = 5.6;
  A->data[5] = 6.7;
  printmat(A);

  Matrix *B = createMatrix(2, 3);
  B->data[0] = 5.5;
  B->data[1] = 6.6;
  B->data[2] = 7.7;
  B->data[3] = 1.2;
  B->data[4] = 2.1;
  B->data[5] = 3.3;
  printmat(B);

  Matrix *C = createMatrix(3, 3);
  matrixmult(A, B, C);
  printmat(C);

  destroyMatrix(A);
  destroyMatrix(B);
  destroyMatrix(C);
  return 0;
}

// your code goes below...


Matrix *createMatrix(int nrows, int ncols)
{
  // fill in the code here
}

void destroyMatrix(Matrix *M)
{
  // fill in the code here
}

void printmat(Matrix *M)
{
  // fill in the code here
  printf("so far printmat does nothing\n");
}

void matrixmult(Matrix *A, Matrix *B, Matrix *C)
{
  // fill in the code here
  printf("so far matrixmult does nothing\n");
}


Solutions


Paul Gribble | Summer 2012
This work is licensed under a Creative Commons Attribution 4.0 International License
Creative Commons License