How do function pointers in C work?

Question

I had some experience lately with function pointers in C.

So going on with the tradition of answering your own questions, I decided to make a small summary of the very basics, for those who need a quick dive-in to the subject.

Answer 1

function pointers are useful in many situations, eg:

• COM objects members are pointer to function ag: This->lpVtbl->AddRef(This); AddRef is a pointer to a function.
• function callback, for example a user defined function to compares two variables to be passed as a callback to a special sort function.
• very useful for plugin implementation and application SDK.

Answer 2

One of my favorite uses for function pointers is as cheap and easy iterators -

#include <stdio.h>
#define MAX_COLORS  256

typedef struct {
    char* name;
    int red;
    int green;
    int blue;
} Color;

Color Colors[MAX_COLORS];


void eachColor (void (*fp)(Color *c)) {
    int i;
    for (i=0; i<MAX_COLORS; i++)
        (*fp)(&Colors[i]);
}

void printColor(Color* c) {
    if (c->name)
        printf("%s = %i,%i,%i\n", c->name, c->red, c->green, c->blue);
}

int main() {
    Colors[0].name="red";
    Colors[0].red=255;
    Colors[1].name="blue";
    Colors[1].blue=255;
    Colors[2].name="black";

    eachColor(printColor);
}

Answer 3

Function pointers in C

Let's start with a basic function which we will be pointing to:

int addInt(int n, int m) {
    return n+m;
}

First thing, let's define a pointer to a function which receives 2 ints and returns an int:

int (*functionPtr)(int,int);

Now we can safely point to our function:

functionPtr = &addInt;

Now that we have a pointer to the function, let's use it:

int sum = (*functionPtr)(2, 3); // sum == 5

Passing the pointer to another function is basically the same:

int add2to3(int (*functionPtr)(int, int)) {
    return (*functionPtr)(2, 3);
}

We can use function pointers in return values as well (try to keep up, it gets messy):

// this is a function called functionFactory which receives parameter n
// and returns a pointer to another function which receives two ints
// and it returns another int
int (*functionFactory(int n))(int, int) {
    printf("Got parameter %d", n);
    int (*functionPtr)(int,int) = &addInt;
    return functionPtr;
}

But it's much nicer to use a typedef:

typedef int (*myFuncDef)(int, int);
// note that the typedef name is indeed myFuncDef

myFuncDef functionFactory(int n) {
    printf("Got parameter %d", n);
    myFuncDef functionPtr = &addInt;
    return functionPtr;
}

Answer 4

Starting from scratch function has Some Memory Address From Where They start executing. In Assembly Language They Are called as (call "function's memory address").Now come back to C If function has a memory address then they can be manipulated by Pointers in C.So By the rules of C

1.First you need to declare a pointer to function 2.Pass the Address of the Desired function

****Note->the functions should be of same type****

This Simple Programme will Illustrate Every Thing.

#include<stdio.h>
void (*print)() ;//Declare a  Function Pointers
void sayhello();//Declare The Function Whose Address is to be passed
                //The Functions should Be of Same Type
int main()
{

 print=sayhello;//Addressof sayhello is assigned to print
 print();//print Does A call To The Function 
 return 0;
}

void sayhello()
{
 printf("\n Hello World");
}

After That lets See How machine Understands Them.Glimpse of machine instruction of the above programme in 32 bit architecture.

The red mark area is showing how the address is being exchanged and storing in eax.Then their is a call instruction on eax. eax contains the desired address of the function

Answer 5

A function pointer is a variable that contains the address of a function. Since it is a pointer variable though with some restricted properties, you can use it pretty much like you would any other pointer variable in data structures.

The only exception I can think of is treating the function pointer as pointing to something other than a single value. Doing pointer arithmetic by incrementing or decrementing a function pointer or adding/subtracting an offset to a function pointer isn't really of any utility as a function pointer only points to a single thing, the entry point of a function.

The size of a function pointer variable, the number of bytes occupied by the variable, may vary depending on the underlying architecture, e.g. x32 or x64 or whatever.

The declaration for a function pointer variable needs to specify the same kind of information as a function declaration in order for the C compiler to do the kinds of checks that it normally does. If you don't specify a parameter list in the declaration/definition of the function pointer, the C compiler will not be able to check the use of parameters. There are cases when this lack of checking can be useful however just remember that a safety net has been removed.

Some examples:

int func (int a, char *pStr);    // declares a function

int (*pFunc)(int a, char *pStr);  // declares or defines a function pointer

int (*pFunc2) ();                 // declares or defines a function pointer, no parameter list specified.

int (*pFunc3) (void);             // declares or defines a function pointer, no arguments.

The first two declararations are somewhat similar in that:

• func is a function that takes an int and a char * and returns an int
• pFunc is a function pointer to which is assigned the address of a function that takes an int and a char * and returns an int

So from the above we could have a source line in which the address of the function func() is assigned to the function pointer variable pFunc as in pFunc = func;.

Notice the syntax used with a function pointer declaration/definition in which parenthesis are used to overcome the natural operator precedence rules.

int *pfunc(int a, char *pStr);    // declares a function that returns int pointer
int (*pFunc)(int a, char *pStr);  // declares a function pointer that returns an int

Several Different Usage Examples

Some examples of usage of a function pointer:

int (*pFunc) (int a, char *pStr);    // declare a simple function pointer variable
int (*pFunc[55])(int a, char *pStr); // declare an array of 55 function pointers
int (**pFunc)(int a, char *pStr);    // declare a pointer to a function pointer variable
struct {                             // declare a struct that contains a function pointer
    int x22;
    int (*pFunc)(int a, char *pStr);
} thing = {0, func};                 // assign values to the struct variable
char * xF (int x, int (*p)(int a, char *pStr));  // declare a function that has a function pointer as an argument
char * (*pxF) (int x, int (*p)(int a, char *pStr));  // declare a function pointer that points to a function that has a function pointer as an argument

You can use variable length parameter lists in the definition of a function pointer.

int sum (int a, int b, ...);
int (*psum)(int a, int b, ...);

Or you can not specify a parameter list at all. This can be useful but it eliminates the opportunity for the C compiler to perform checks on the argument list provided.

int  sum ();      // nothing specified in the argument list so could be anything or nothing
int (*psum)();
int  sum2(void);  // void specified in the argument list so no parameters when calling this function
int (*psum2)(void);

C style Casts

You can use C style casts with function pointers. However be aware that a C compiler may be lax about checks or provide warnings rather than errors.

int sum (int a, char *b);
int (*psplsum) (int a, int b);
psplsum = sum;               // generates a compiler warning
psplsum = (int (*)(int a, int b)) sum;   // no compiler warning, cast to function pointer
psplsum = (int *(int a, int b)) sum;     // compiler error of bad cast generated, parenthesis are required.

Compare Function Pointer to Equality

You can check that a function pointer is equal to a particular function address using an if statement though I am not sure how useful that would be. Other comparison operators would seem to have even less utility.

static int func1(int a, int b) {
    return a + b;
}

static int func2(int a, int b, char *c) {
    return c[0] + a + b;
}

static int func3(int a, int b, char *x) {
    return a + b;
}

static char *func4(int a, int b, char *c, int (*p)())
{
    if (p == func1) {
        p(a, b);
    }
    else if (p == func2) {
        p(a, b, c);      // warning C4047: '==': 'int (__cdecl *)()' differs in levels of indirection from 'char *(__cdecl *)(int,int,char *)'
    } else if (p == func3) {
        p(a, b, c);
    }
    return c;
}

An Array of Function Pointers

And if you want to have an array of function pointers each of the elements of which the argument list has differences then you can define a function pointer with the argument list unspecified (not void which means no arguments but just unspecified) something like the following though you may see warnings from the C compiler. This also works for a function pointer parameter to a function:

int(*p[])() = {       // an array of function pointers
    func1, func2, func3
};
int(**pp)();          // a pointer to a function pointer


p[0](a, b);
p[1](a, b, 0);
p[2](a, b);      // oops, left off the last argument but it compiles anyway.

func4(a, b, 0, func1);
func4(a, b, 0, func2);  // warning C4047: 'function': 'int (__cdecl *)()' differs in levels of indirection from 'char *(__cdecl *)(int,int,char *)'
func4(a, b, 0, func3);

    // iterate over the array elements using an array index
for (i = 0; i < sizeof(p) / sizeof(p[0]); i++) {
    func4(a, b, 0, p[i]);
}
    // iterate over the array elements using a pointer
for (pp = p; pp < p + sizeof(p)/sizeof(p[0]); pp++) {
    (*pp)(a, b, 0);          // pointer to a function pointer so must dereference it.
    func4(a, b, 0, *pp);     // pointer to a function pointer so must dereference it.
}

C style namespace Using Global struct with Function Pointers

You can use the static keyword to specify a function whose name is file scope and then assign this to a global variable as a way of providing something similar to the namespace functionality of C++.

In a header file define a struct that will be our namespace along with a global variable that uses it.

typedef struct {
   int (*func1) (int a, int b);             // pointer to function that returns an int
   char *(*func2) (int a, int b, char *c);  // pointer to function that returns a pointer
} FuncThings;

extern const FuncThings FuncThingsGlobal;

Then in the C source file:

#include "header.h"

// the function names used with these static functions do not need to be the
// same as the struct member names. It's just helpful if they are when trying
// to search for them.
// the static keyword ensures these names are file scope only and not visible
// outside of the file.
static int func1 (int a, int b)
{
    return a + b;
}

static char *func2 (int a, int b, char *c)
{
    c[0] = a % 100; c[1] = b % 50;
    return c;
}

const FuncThings FuncThingsGlobal = {func1, func2};

This would then be used by specifying the complete name of global struct variable and member name to access the function. The const modifier is used on the global so that it can not be changed by accident.

int abcd = FuncThingsGlobal.func1 (a, b);

Application Areas of Function Pointers

A DLL library component could do something similar to the C style namespace approach in which a particular library interface is requested from a factory method in a library interface which supports the creation of a struct containing function pointers.. This library interface loads the requested DLL version, creates a struct with the necessary function pointers, and then returns the struct to the requesting caller for use.

typedef struct {
    HMODULE  hModule;
    int (*Func1)();
    int (*Func2)();
    int(*Func3)(int a, int b);
} LibraryFuncStruct;

int  LoadLibraryFunc LPCTSTR  dllFileName, LibraryFuncStruct *pStruct)
{
    int  retStatus = 0;   // default is an error detected

    pStruct->hModule = LoadLibrary (dllFileName);
    if (pStruct->hModule) {
        pStruct->Func1 = (int (*)()) GetProcAddress (pStruct->hModule, "Func1");
        pStruct->Func2 = (int (*)()) GetProcAddress (pStruct->hModule, "Func2");
        pStruct->Func3 = (int (*)(int a, int b)) GetProcAddress(pStruct->hModule, "Func3");
        retStatus = 1;
    }

    return retStatus;
}

void FreeLibraryFunc (LibraryFuncStruct *pStruct)
{
    if (pStruct->hModule) FreeLibrary (pStruct->hModule);
    pStruct->hModule = 0;
}

and this could be used as in:

LibraryFuncStruct myLib = {0};
LoadLibraryFunc (L"library.dll", &myLib);
//  ....
myLib.Func1();
//  ....
FreeLibraryFunc (&myLib);

The same approach can be used to define an abstract hardware layer for code that uses a particular model of the underlying hardware. Function pointers are filled in with hardware specific functions by a factory to provide the hardware specific functionality that implements functions specified in the abstract hardware model. This can be used to provide an abstract hardware layer used by software which calls a factory function in order to get the specific hardware function interface then uses the function pointers provided to perform actions for the underlying hardware without needing to know implementation details about the specific target.

Function Pointers to create Delegates, Handlers, and Callbacks

You can use function pointers as a way to delegate some task or functionality. The classic example in C is the comparison delegate function pointer used with the Standard C library functions qsort() and bsearch() to provide the collation order for sorting a list of items or performing a binary search over a sorted list of items. The comparison function delegate specifies the collation algorithm used in the sort or the binary search.

Another use is similar to applying an algorithm to a C++ Standard Template Library container.

void * ApplyAlgorithm (void *pArray, size_t sizeItem, size_t nItems, int (*p)(void *)) {
    unsigned char *pList = pArray;
    unsigned char *pListEnd = pList + nItems * sizeItem;
    for ( ; pList < pListEnd; pList += sizeItem) {
        p (pList);
    }

    return pArray;
}

int pIncrement(int *pI) {
    (*pI)++;

    return 1;
}

void * ApplyFold(void *pArray, size_t sizeItem, size_t nItems, void * pResult, int(*p)(void *, void *)) {
    unsigned char *pList = pArray;
    unsigned char *pListEnd = pList + nItems * sizeItem;
    for (; pList < pListEnd; pList += sizeItem) {
        p(pList, pResult);
    }

    return pArray;
}

int pSummation(int *pI, int *pSum) {
    (*pSum) += *pI;

    return 1;
}

// source code and then lets use our function.
int intList[30] = { 0 }, iSum = 0;

ApplyAlgorithm(intList, sizeof(int), sizeof(intList) / sizeof(intList[0]), pIncrement);
ApplyFold(intList, sizeof(int), sizeof(intList) / sizeof(intList[0]), &iSum, pSummation);

Another example is with GUI source code in which a handler for a particular event is registered by providing a function pointer which is actually called when the event happens. The Microsoft MFC framework with its message maps uses something similar to handle Windows messages that are delivered to a window or thread.

Asynchronous functions that require a callback are similar to an event handler. The user of the asynchronous function calls the asynchronous function to start some action and provides a function pointer which the asynchronous function will call once the action is complete. In this case the event is the asynchronous function completing its task.

Answer 6

Another good use for function pointers:
Switching between versions painlessly

They're very handy to use for when you want different functions at different times, or different phases of development. For instance, I'm developing an application on a host computer that has a console, but the final release of the software will be put on an Avnet ZedBoard (which has ports for displays and consoles, but they are not needed/wanted for the final release). So during development, I will use printf to view status and error messages, but when I'm done, I don't want anything printed. Here's what I've done:

version.h

// First, undefine all macros associated with version.h
#undef DEBUG_VERSION
#undef RELEASE_VERSION
#undef INVALID_VERSION


// Define which version we want to use
#define DEBUG_VERSION       // The current version
// #define RELEASE_VERSION  // To be uncommented when finished debugging

#ifndef __VERSION_H_      /* prevent circular inclusions */
    #define __VERSION_H_  /* by using protection macros */
    void board_init();
    void noprintf(const char *c, ...); // mimic the printf prototype
#endif

// Mimics the printf function prototype. This is what I'll actually 
// use to print stuff to the screen
void (* zprintf)(const char*, ...); 

// If debug version, use printf
#ifdef DEBUG_VERSION
    #include <stdio.h>
#endif

// If both debug and release version, error
#ifdef DEBUG_VERSION
#ifdef RELEASE_VERSION
    #define INVALID_VERSION
#endif
#endif

// If neither debug or release version, error
#ifndef DEBUG_VERSION
#ifndef RELEASE_VERSION
    #define INVALID_VERSION
#endif
#endif

#ifdef INVALID_VERSION
    // Won't allow compilation without a valid version define
    #error "Invalid version definition"
#endif

In version.c I will define the 2 function prototypes present in version.h

version.c

#include "version.h"

/*****************************************************************************/
/**
* @name board_init
*
* Sets up the application based on the version type defined in version.h.
* Includes allowing or prohibiting printing to STDOUT.
*
* MUST BE CALLED FIRST THING IN MAIN
*
* @return    None
*
*****************************************************************************/
void board_init()
{
    // Assign the print function to the correct function pointer
    #ifdef DEBUG_VERSION
        zprintf = &printf;
    #else
        // Defined below this function
        zprintf = &noprintf;
    #endif
}

/*****************************************************************************/
/**
* @name noprintf
*
* simply returns with no actions performed
*
* @return   None
*
*****************************************************************************/
void noprintf(const char* c, ...)
{
    return;
}

Notice how the function pointer is prototyped in version.h as

void (* zprintf)(const char *, ...);

When it is referenced in the application, it will start executing wherever it is pointing, which has yet to be defined.

In version.c, notice in the board_init()function where zprintf is assigned a unique function (whose function signature matches) depending on the version that is defined in version.h

zprintf = &printf; zprintf calls printf for debugging purposes

or

zprintf = &noprint; zprintf just returns and will not run unnecessary code

Running the code will look like this:

mainProg.c

#include "version.h"
#include <stdlib.h>
int main()
{
    // Must run board_init(), which assigns the function
    // pointer to an actual function
    board_init();

    void *ptr = malloc(100); // Allocate 100 bytes of memory
    // malloc returns NULL if unable to allocate the memory.

    if (ptr == NULL)
    {
        zprintf("Unable to allocate memory\n");
        return 1;
    }

    // Other things to do...
    return 0;
}

The above code will use printf if in debug mode, or do nothing if in release mode. This is much easier than going through the entire project and commenting out or deleting code. All that I need to do is change the version in version.h and the code will do the rest!