Passing an array as an argument in C++

I'm writing a merge sort function, and right now I am just using a test case array (there is no input - this is static, for now). I don't know how to pass an array as an argument. Here is my code right now:

//merge sort first attempt

#include <iostream>
#include <algorithm>
#include <vector>

int mergeSort(int[]);
int main()
{
int originalarray[] = {1, 3, 5, 7, 9, 2, 4, 6, 8, 10};
mergeSort(originalarray[]);
}

int mergeSort(int[] originalarray)
{
int num = (sizeof(originalarray)/sizeof(int));
std::vector<int> original(num);

if (num > 2) {
    return num;
}

// Fill the array using the elements of originalarray
// This is just for demonstration, normally original will be a parameter,
// so you won't be filling it up with anything.
std::copy(originalarray, originalarray + num, original.begin());

// Create farray and sarray of the appropriate size
std::vector<int> farray(num / 2);
std::vector<int> sarray(num - farray.size());

// Fill those using elements from original
std::copy(original.begin(), original.begin() + farray.size(), farray.begin());
std::copy(original.begin() + farray.size(), original.end(), sarray.begin());

mergeSort(farray);
mergeSort(sarray);

}

Note that this mergeSort function is not functional, as I have not figured out how to merge them yet (that's my assignment). I would like to get my two vectors sorted before I deal with that, and I can't compile this because of my need to pass an array as an argument. I don't understand pointers, so if that is the solution, my excuse is ignorance. I'm learning programming right now, with C++ as a first language, and only have a basic grasp of the language's features. Thanks for the help.

From stackoverflow

• You should not use sizeof(originalarray)/sizeof(int) like that. It'll only work for statically declared arrays (the size is known at compile time). You have to pass the size along with it. Why don't you just make a vector out of the array and pass it instead?

  Side Note: As a rule of thumb, always note that sizeof will be translated at compile time. So there's no way it could know the size of the array passed as an argument.

  Hooked : I don't know what you mean by that. Should I just std::copy all the elements into a vector and pass that? How do you use sizeof() with a vector?

  Mehrdad Afshari : You don't use sizeof on vector. v.size() will get the size. You should either do that, or if you want to pass an array, pass the size as another parameter.

• • When you pass arrays to functions, they decay to pointers to the first element of the array, the notation notwithstanding. So, your sizeof doesnot work as expected.

  • When you pass in an array, it is best to pass in the array size, so that you know where to stop. Add it as an additional parameter.

• I see you include <vector>. I suggest you do away with all uses of arrays and only use the vector class. You can see examples of how to use STL containers such as vector here.

  Dan : I strongly concur.

• Jut to extend this a bit, remember that C++ arrays are exactly C arrays. So all you have is the address of a piece of memory that purports (with no guarantees) to be an array of somethings.

  Update

  Okay, we'll expand a little more.

  C (and therefore C++) doesn't really have "arrays" as such. All it has are addresses, pointers. So when you make something an "array", what really happens is you tell the compiler that some variable represents an address.

  It's useful to make a distinction in C between a declaration and a definition. In a declaration, you're simply giving something a name and a type; in a definition, you actually allocate space.

  So, if we start off by definiing an array like

  int ar[100];
  

  that means we're telling the compiler we want space for 100 int's, we want it to all be allocated in one chunk, and we're going to use the name ar for it. The sizeof operator gives the number of bytes used by a type or an object, so our array ar will take up 100×sizeof(int) bytes. On most machines, that will be 400 bytes, but it varies from machine to machine.

  If we define a variable

  int * ar_p;   // using '_p' as a reminder this is a pointer
  

  we're defining space for a variable that will contain an address. Its size will be sizeof(int*), which will usually be either 4 or 8, but on some machines could be anything from 2 to 16 on some machines you're unlikely to run into soon.

  The name of the array is `ar. The compiler converts that name into an address, so we can save that address with

  ar_p = ar ;
  

  Now, let's say for convenience that our array ar happened to be starting at location 1000 in memory.

  That name ar` does not have any space allocated to it; it's like a constant, a number. So, you can't reverse that assignment

  ap = ar_p ;     // THIS WON'T WORK
  

  for the same reason you couldn't say

  1000 = ar_p ;   // THIS WON'T WORK EITHER
  

  ie, you can't change the value of 1000. (Back in early versions of FORTRAN, this trick would work, for complicated reasons. It was a mistake. You've never lived until you've tried to debug a program in which the value of "2" is 3.)

  Arrays in C are always zero-based, that is, the first index is always zero. Any other indices are just addresses computed using the index. So, ar[0] is just the address 1000 plus 0 bytes of offset, or 1000. ar[1] is 1000 plus 1 times the size of an int, so the next int over. And in fact, this is always true in C.

  This is called an array reference.

  When we use the syntax *ar_p we're telling the compiler to get the thing AT the address contained in ar_p. `.

  This is called dereferencing a pointer.

  If we say

  ar_p = ar;
  

  then *ar_p and ar[0] refer to the same thing.

  When we say ar[0] we're telling the compiler we want the thing at the address 0 bytes from ar. ar[1] is the address one int, or 4 bytes, from ar. So, *(ar_p+3) refers to the same thing as ar[3]. (We need the parentheses because we want to add 3 to the address first and then look at the contents. *ar_p+3 would get the contents pointed to by ap_p first, and then add 3 to those.

  The thing is, C doesn't know, or much care, how big the array really is. If I come along and do ar[365], the compiler will happily generate code to look in the cell 1000+(365×sizeof(int)). If that's in your array, fine, but if it's just random memory, that's fine too. C doesn't care.

  (Remember C comes from the phone company. "We don't care; we don't have to. We're the Phone Company.")

  So, now, we know some rules, which I've moved down here. Read "≡" as "is equivalent to" or "is the same as".

  What you can depend on:

  • foo(TYPE t[]) ≡ foo(TYPE * t)

  Since C doesn't know a difference between pointers and arrays, you can declare either one. When you define a function, you can write

  void foo(int[] ar){
  

  or

  void foo(int* ar){
  

  and get exactly the same effect.

  • t[i] ≡ *(t+i)

  This was above. Anywhere you might write ar[i], you can replace it with *(ar+i). (There's actually a weird side case that breaks this, but you won't run into it as a beginner.)

  • where TYPE *t, (t+i) will equal the address at t plus i*sizeof(TYPE)

  Explained this above as well. When you index into an array, like ar[42], it means you want the 42nd whatever over from the start address. So, if you're using int, then you need to move over 42 times however wide an int is, which is to say sizeof(int).

  Now, that's all C, and since C++ is defined as a "kind of" C, it all holds for C++ as well. EXCEPT

  • unless TYPE is a user defined type that overloads operator[] and operator*.

  in C++, you can decide you want to define a new type that acts just like any other type, but you can change the way the language does specific things. So, a programmer can decide to "overload" -- ie, replace -- the default behavior of the array reference and pointer dereference operators with something of their own devising. As a beginner, you shouldn't be confronted with that soon, but you should be aware of it.

  Hooked : I think you overestimate my knowledge, because I have know idea what you just said. I'm currently learning programming right now, and am not onto pointers or the OOP parts of C++ yet.

  Charlie Martin : Ah. Okay, I'll add some.

  AlexW : That's a really good breakdown of arrays in C++. I'm glad I use .NET/Java! :)

  Charlie Martin : Thanks for the kind words, Alex, but actually understanding this would make a big difference to your understanding of Java/C# as well. For example, what to you get when you say Object a = new Object(); in Java. What is 'a' really?

• In addition to all the answers above, you might also want to check out the Q&As on arrays from c-faq.com: http://c-faq.com/aryptr/index.html

• Unfortunately, it's very hard to do exactly what you want to do in C or C++. You can pass around a fixed-size array like this:

  int mergeSort(int originalarray[20])
  {
      // do something
  }
  

  However, your array's size is not defined by a number, it's defined by the number of elements in initialization list.

  The thing to do in your case (even though it's really a wrong thing to do) is to do it in two steps:

  int originalarray[] = {1, 3, 5, 7, 9, 2, 4, 6, 8, 10};
  const size_t arraySize = sizeof originalarray / sizeof originalarray[0];
  int mergeSort(int array[arraySize])
  {
      // do something
  }
  

  Too bad it will not do what you need done: passing the array to a function like this makes a copy of the array, and the point of sorting would be to change the original array.

  In truth, you cannot go any further without understanding the concept of "pointer".

  The function you need to develop really should be like this:

  int originalarray[] = {1, 3, 5, 7, 9, 2, 4, 6, 8, 10};
  const size_t arraySize = sizeof originalarray / sizeof originalarray[0];
  
  int mergeSort(int *array, const size_t size)
  {
      // do something
  }
  
  mergeSort(&(originalArray[0]), arraySize);
  

  In other words, you pass a pointer to first element, and the number of elements.

  Alternatively, you can deal with vectors. Vector encapsulates the same two things (pointer to first element and size) in a single entity called "object". Plus, it manages memory for you, so you can extend the number of elements as you need. This is the C++ way. Too bad you can't initialize a vector with {...} like you can an array.

• Looks like you're using both dynamically allocated arrays and vectors, when I believe just using std::vector will be enough.

  First, let your input array be changed to a std::vector, and fill it with your input data.

  int main()
  {
     std::vector<int> originalarray;
     for (int data = 1; data <= 10; data++)
     {
        originalarray.push_back(data);
     }
     mergeSort(originaldata);
  }
  

  Now it's important to declare your mergesort function to take a reference to a std::vector.

  int mergeSort(std::vector<int>& originalarray)
  {
     // The rest of your code, note that now you are passing 
     // in your array for sorting, so you can continue with your code to split
     // the vector into farray and sarray
  
     // then call sort on your halves.
     mergeSort(farray);
     mergeSort(sarray);
  
     // I'm guessing at this point you'd write code to combine your farray sarray, and
     // put it back into originalarray...don't forget to clear original array first!
  }
  

  Just a note, looks like you're not doing an inplace sort, so expect your sort to take a while since you're copying out a lot of data.