Passing a sub-range of a vector by reference

Hello,

Suppose I have a stl vector of ints, and I want to pass a sub-range of that vector as an argument to a function. One way of doing that would be to copy the sub-range, and then pass that copy by reference, as follows:

Code:

---

#include <vector>

using namespace std;

int MyFunction(vector<int> &a_vector)
{
    // Do something

    return 0;
};

int main()
{
    // Create a vector and assign its elements;

    vector<int> my_vector(4);
    my_vector[0] = 10;
    my_vector[1] = 20;
    my_vector[2] = 50;
    my_vector[3] = 100;

    // Create a sub-vector and assign its elements

    int sub_vector_length = 2;
    vector<int> sub_vector(sub_vector_length);
    for (int i = 0; i < sub_vector_length; i++)
    {
        sub_vector[i] = my_vector[i];
    }

    // Pass the sub-vector by reference to MyFunction
    MyFunction(sub_vector);

    return 0;
}

---

However, it can be time-consuming to copy the elements between the two vectors if my_vector is large. So I'm wondering if it is possible to directly pass a sub-range of my_vector by reference, without having to create a new vector and manually copy over all of the relevant elements?

Thanks!

Ed.

Quote:

---

Originally Posted by ejohns85

However, it can be time-consuming to copy the elements between the two vectors if my_vector is large. So I'm wondering if it is possible to directly pass a sub-range of my_vector by reference, without having to create a new vector and manually copy over all of the relevant elements?

---

Just pass a pointer to the first item in the vector you want to use in the function. A vector is a wrapper for a contiguous, dynamic array of T, so there is no need to be copying anything.

Code:

---

int MyFunction(int *a_vector)
{
    // Do something

    return 0;
};
//...
// Pass the sub-vector by reference to MyFunction
MyFunction(&my_vector[2]);

---

Regards,

Paul McKenzie

If you can change the function, you can design it similar to the way the algorithms
work: pass a start iterator and an end iterator (one past last element).

Quote:

---

Originally Posted by ejohns85

However, it can be time-consuming to copy the elements between the two vectors if my_vector is large.

---

Also, there is really no need to write the loop yourself.

Code:

---

int sub_vector_length = 2;
vector<int> sub_vector(myvector.begin(), myvector.begin() + sub_vector_length);

---

If the compiler's optimizer is a good one, and the compiler recognizes that "int" is a POD, then the copying of a vector of POD type to another could be optimized by issuing a memcpy(). So the copying of even large areas of memory for POD types becomes almost a moot point, as memcpy() is usually optimized to a few assembly language calls.

Regards,

Paul McKenzie

Quote:

---

Originally Posted by ejohns85

So I'm wondering if it is possible to directly pass a sub-range of my_vector by reference, without having to create a new vector and manually copy over all of the relevant elements?

---

You could pass a start-of-range iterator and a range length to the function. The function can then use the iterator for either sequential or random access to the elements within the range.

More ambitious would be to introduce a class called say Subrange which is used to access a vector subrange while hiding the details. A Subrange object would then be passed to the function. Clean, simple, safe and efficient (at least potentially :)).