new versus *malloc

quote:

---

Original post by Omaha
...if the malloc family is obsolete now that new has come to town with easy dynamic memory allocation for all.

---

type *ptr;
// allocate the original array
ptr = new type[num];
// reallocate array:
type *tmp = new type[new_num];
// copy items from ptr to tmp
for(int k = 0; k < ptr_size; ++k)
  tmp[k] = ptr[k];
// delete existing array
delete ptr;
// switch pointers
ptr = tmp;
 

quote:

---

Original post by Oluseyi
I''ve never used realloc, and never found a reason to.

---

quote:

---

Original post by Beer Hunter
I can only imagine one reason not to have a realloc equivalent in c++, and it''s not a very good reason at all.

---

quote:

---

Original post by Stoffel
As far as I understand it, here's the logic for realloc:
1) check through the memory manager to see if there's enough room after the requested allocation for the new size--potential an expensive operation (?)
2) if there is, update the allocation table
3) if not, allocate new memory size, memcpy, and deallocate old memory.

The reason I thought it was ditched was that step #1 almost always failed, so why fool a user into thinking realloc was faster when the user can just skip to step 3 always and have a faster program?

---

  
struct CHUNK {
  CHUNK   *next;
  size_t   size;
  size_t   maxSize;
};
void *realloc (void *ptr, size_t size)
{
  CHUNK *chunk = ptr - sizeof(CHUNK);
  if (size <= chunk->maxSize) {
    // We don't have to move the memory (step 2).

  } else {
    // We have to move the memory (step 3).

  }
}