Char-array vs Char-pointer

I was wondering if there is some advantage of using a char-array over a char-pointer? From were I can see char-array are static and in most cases
consumes more memory than pointers.

Code:

---

#include <iostream>


using namespace std;

int main()
{

        char name[5] = "Eric";
        char *pname = "Eric";
       
        cout << "char-arrays name is " << name << endl;
        cout << "char-pointers name is " << pname << endl << endl;

        //name = "Alexander";  //Out if bounds
        pname = "Alexander";  //No problemo

        cout << "After new assignment..." << endl;
       
        cout << "char-arrays name is " << name << endl;
        cout << "char-pointers name is " << pname << endl << endl;

        cout << "And now the size of them" << endl;
        cout << "char-arrays size is " << sizeof(name) << " bytes" <<endl;
        cout << "char-pointers size is " << sizeof(*pname) << endl << endl;
        //Don´t really sure which of them to use, but I guessed that it has to be
        //the object it points to
        //cout << "char-pointers size is " << sizeof(pname) << endl << endl;

return 0;
}

---

Why use a char-array?????

>Why use a char-array?????

Because it'll comsume exactly the same amout of memory as a literal string and be guaranteed to be modifyable.

You're misunderstanding a char pointer. It doesn't contain the whole string in its first byte.

char* ptr = "Enmeduranki";

the sizeof ptr will be one (that's the sizeof the ptr, not the string). The sizeof the whole string will be 12.

"Enmeduranki" is a string of a certain length, no matter where it is stored it'll consume the same amount of memory (assuming that it is stored as the same data type (ANSI and UNICODE) in each section).

k, both consumes equal memory.
Forgot to take the sum of sizeof(pname[0])+sizeof(pname[1)+...sizeof(pname[4]).

But still isn´t a char-pointer more flexible. You can assign it (multiple times) with strings of different length which you can´t do it with a char-array. Have you once defined a char-array of 5 element there is no way to "extend it". Isn´t this an advantage for using a char-pointers insteed of char-arrays????

>You can assign it

You're not really assigning anything though (apart from the intial address). It's still the same string. If you need a discreet copy then you'll need some memory.

>Have you once defined a char-array of 5 element there is no way to "extend it". <

That's how computers work. You ask for a certain amount of memory, and that's what you get back. If you need to extend it then you need more memory. If you're extending literal strings then you stand a good chance of walking into memory that doesn't know that it's allocated, or memory that is used for something else (which could end in tears the next time you need memory).

>Isn´t this an advantage for using a char-pointers insteed of char-arrays????<

The only advantage in the present enviroments is when you want a constant string. This may be faster if you're using alot of strings or your brain operates at ~1GHz.

in my opinion, char pointers are a lot more flexible, because say you want, during the program, to assign the pointer to a string of an unknown length, you can just use malloc() or globalalloc() or whichever you want, and it gives you only as much memory as you need. to use the same pointer again, just free() it, then use malloc() again.

however, with char arrays, you must set the size when you declare the char[], and you can never change it.

here is a quick example of the good way of doing it:

Code:

---

char *calloc(long lSize)
{
    char *tmpPtr;
   
    if ( (tmpPtr=(char*)malloc(lSize))==NULL)
        exit(1);

    return tmpPtr;
}

---

What that code does is returns a pointer to a char of lSize bytes length. i find that this is the most efficient way to do chars.

Quote:

---

char *pname = "Eric";

pname = "Alexander"; //No problemo

---

No, that's a BIG problemo. You got lucky, and the program didn't crash, but you should not use pointers this way. When you point a pointer at a string literal like that, don't attempt to modify it or even repoint it. It's const, for all practical purposes. This is so important to understand.

One last thing, and that's about calling malloc and other memory allocation functions. Believe it or not, they can lead to A LOT of memory fragmentation. This is not a problem for small programs, but when you start playing with more memory, you will one day get a big byte in the *** when malloc returns NULL! So what's my point, you say? At some point you should consider large scale allocation and management schema. C++ is perfect for the job because of it's many useful constructs. So how about a simple example?

For a given program, guestimate how much total memory you'll need. Allocate that much ( but perhaps a bit more, right?) straight from a special class object.

Code:

---



class Memory {

unsigned char *heap;

unsigned char *current;

double heapsize;

double remaining;

bool bailout;

public:

Memory( double initial_size, bool yes_or_no = true )
 :heapsize(initial_size), remaining( initial_size ),
  current(heap), bailout( yes_or_no )
 { 
  heap = (unsigned char*)malloc(sizeof(unsigned char));
 
  if( heap == NULL )
  {
    if(bailout)
    {
      cerr << "Not Enough Memory!!!";
      abort();   
    }                //...otherwise...
    heap = NULL;
    heapsize = remaining = 0;
  }

  current = heap;   
 
  memset(heap, 0, heapsize);
 }


void * malloc( int amount )
 {
  if( amount > remaining + 1)
  {
    if( remaining == 0 ) return NULL;

    amount = remaining;   
  }
  amount++;  //...to advance to the next block...
 
  void * chunk = (void*)current;

  current += amount;
 
  remaining -= amount;

  return chunk; 
 }


char * cMalloc( int amount )
 {
  return (char*) Memory::malloc( amount );
 }

int * iMalloc( int amount )
 {
  return (int*) Memory::malloc( amount * sizoef(int) );
 }

~Memory()
 {
  if( heap != NULL )
  {
    free(heap);
    heap = NULL;
  }
 }
};

---

In main, the calling would be familar:

Code:

---



Memory mem( 1000000 ); //...get a MB...

int main()
{
 char * str = mem.cMalloc(100);
 
 int * array = mem.iMalloc(200);
 
 Employee *emps = (Employee*)mem.malloc( sizeof(Employee) * 10 );
 //...an array of 10 Employees...

return 0;
}

//...destuctor is automatically called, hey, that's garbage collection!

---

So why go to all the trouble? Simple. First, we can ensure we have enough "startup memory" before going to all the trouble of setting up the program, only to crash 15 minutes later because the system says "no", for instance. But most importantly, we can very easily extend this class to do defragmentation too (automatically, of course, perhaps in our own "free()" method). Third, we have an automatic garbage collector, so we don't have to free up the memory when the program is done ( or crashes for some other reason, etc).

Instead of cMalloc and iMalloc why not create a function like this:

Code:

---

template<typename T> T* Malloc(int amount)
{
  return reinterpret_cast<T*>(Memory::malloc(amount * sizeof(T)));
}


//Then call it like this:

int* data = memory.Malloc<int>(23);

---

Another way is to override the global operator new:

Code:

---

Memory mem; //global object

//override global new operaotr
void* operator new(size_t size)
{
    //use global memory object
    return mem.Malloc(size);
}

int main()
{
  //init memory;
  mem.init(1000000);

  //Very convenient use of the memory class
  int* mynumber = new int;
}

---

Note that the new[] operator must be overrided as well

quoted by Sebastiani

Quote:

---

When you point a pointer at a string literal like that, don't attempt to modify it or even repoint it. It's const, for all practical purposes. This is so important to understand.

---

Does that mean that is is dangerous to assaign pname = "John" (or a string less than 5 in lenght) so that we overwrite memory?And if it is so dangerous why doesn´t compiler warn us???

Not exactly. The main thing is, you are creating strings on the stack that you take no responsibility for. Look at this loop:

while(1)
{
ptr = "Hello World!";
}

For each iteration, the pointer is assigned to a brand new string literal, and all connections to the previous string are lost. Yet the loop looks inocuous. That is why I say "bad practice". Also, let's say you do:

char *ptr = "Hello World";

ptr[0] = 's';

This will probably crash your computer! You do not own the memory and thus cannot manipulate it as such.

So the use of ponters to "anonymous" literals should be avoided unless necessary, and to remind yourself, make it const to boot.

Also, I want to add that the previous poster did an excellent job of extending the functionality of the memory class, one by templatizing the class, and the second by overloading "new", which was exactly what I was hoping someone would propose. So good job, Sang-Drax!

Quote:

---

Originally posted by Sebastiani
So good job, Sang-Drax!

---

Thanks :D

But I doubt that my operator overloding will work.

The new operator can be called before main, for example in a class constructor or something.
If it's called before main(), the Mem object will be uninitialized.

This code can proabably fix it:

Code:

---

void* Memory::Malloc(sizet_t size)
{
  if (!initialized)
    return malloc(size); //standard way
  else
  {
    //normal alloc code
  }
}

---

For example, if you include <iostream> in VC++, the operator new will be called before main().

This is going more and more away from the original topic, perhaps a new one should be created?

Hey that's a great idea...how do we do that again??

....also...

Perhaps you could elaborate on your last post?