Recursive Implementation of Merge Sort C++

Question

I am currently reading an algorithms book and I read on the insertion and merge sorts. I implemented the merge sort in C++, The first function MERGE is responsible for the merging of the two subarrays part and the function MERGE_SORT is the actual algorithm. I am interested to know whether or not my code has a good readability factor and if It can be optimized further or not?

#include <iostream>
#include <vector>
#include <limits>
using namespace std;
#define itr vector<int>::iterator
void MERGE(vector<int>& vec,itr left, itr mid, itr right){
 //Make vector left with elements from left to mid inclusive
 vector<int> left_vec;
 for (itr i=left; i!=vec.end()&&i<=mid;i++){
 left_vec.push_back(*i);
 }
 left_vec.push_back(numeric_limits<int>::max());//sentinel card
 //make vector right with elements from mid+1 to right inclusive
 vector<int> right_vec;
 for (itr i=mid+1; i!=vec.end() && i<=right; i++){
 right_vec.push_back(*i);
 }
 right_vec.push_back(numeric_limits<int>::max());//sentinel card
 //Now add them in a sorted manner to vector from left to right
 itr l=left_vec.begin(); itr r=right_vec.begin();
 for (itr vec_itr=left;vec_itr!=vec.end()&&vec_itr<=right;vec_itr++){
 if (*l<=*r){
 *vec_itr=*l;
 l++;
 }
 else{
 *vec_itr = *r;
 r++;
 }
 }
}
void MERGE_SORT(vector<int>& vec, itr left, itr right){
 if(left<right){
 itr mid = left + (right-left)/2;
 MERGE_SORT(vec,left,mid);
 MERGE_SORT(vec,mid+1,right);
 MERGE(vec, left,mid,right);
 }
 return;
}
int main() {
 int a[] = {10,8,9,7,6,3,3,9,15,4,3,2,1};
 vector<int> vec(a,a+13);
 MERGE_SORT(vec,vec.begin(),vec.end());
 for (itr i=vec.begin();i!=vec.end();i++)cout << *i << " ";
 return 0;
}

Answer 1

First off, I'd use typedef rather than #define.

You don't really need to reference vec.end(); you know mid <= right <= vec.end() and the fact that you're copying the same number of elements back in as you copied out, means you don't need to compare vec_itr to vec.end(). What this really means is that you don't need the vec parameter to MERGE.

vector<T> has a constructor that copies from an iterator. Your push_back for loop could be simplified to

vector<int> left_vec(left, mid);
vector<int> right_vec(mid, right);

The push_back of max() as a sentinel is not strictly needed. Instead, you can just loop until you hit the end of either left_vec or right_vec and then copy the remainder of the vectors to the left to right range.

Your variable names might be easier to understand.

typedef itr vector<int>::iterator;
void MERGE(itr left, itr mid, itr right){
 //Make vector left with elements from left to mid-1 inclusive
 vector<int> left(left, mid);
 //make vector right with elements from mid to right-1 inclusive
 vector<int> right(mid, right);
 // Now concurrently loop over the left and right vectors
 // and add the values back to left to right. Stop when 
 // either iterator runs out of elements.
 itr left_src=left.begin();
 itr right_src=right.begin();
 itr dest = left;
 while (left_src != left.end() && right_src != right.end()) {
 // pick the iterator with the smallest current value
 itr& least_src = (*left_src < *right_src) ? left_src : right_src;
 *dest = *least_src;
 least_src ++;
 dest ++;
 }
 // copy remainder of left (maybe 0 elements)
 copy(left_src, left.end(), dest);
 // copy remainder of right (maybe 0 elements)
 copy(right_src, right.end(), dest);
}