3-way mergesort stackoverflow error in c++ - c++

Question

hi guys can anyone tell me what's wrong with my 3-way mergesort code?the code I wrote can only sort 4 numbers if you give it more than 4 numbers(by changing size) it ends up with stack overflow error,here is the code:
#include "stdafx.h"
#include <iostream>
#include <vector>
using namespace std;
const int size=4;
vector <int> s(size);
void merge(int,int,int);
void mergesort(int,int);
int main(){
for(int i=0;i<size;i++){
cout<<"enter number "<<i+1<<":";
cin>>s.at(i);
}
system("CLS");
cout<<"here are the unsorted numbers:\n";//prints the input values so U can see'em
for(int j=0;j<size;j++)
cout<<s.at(j)<<".";
mergesort(0,size-1);//calls mergesort
cout<<"\nhere are the sorted numbers:\n";
for(int j=0;j<size;j++)
cout<<s.at(j)<<".";
cin.get();
cin.get();
return 0;
}
void merge(int low,int one_third,int high){
int i=low;
int j=one_third+1;
int k=0;
int length=(high-low)+1;
vector <int> u(length,0);
if(k<length){
while((i<=one_third)&&(j<=high)){
if(s.at(i)<=s.at(j)){
u.at(k)=s.at(i);
i++;
k++;
}//end for
else{
u.at(k)=s.at(j);
j++;
k++;
}//end elseif
}//end while
if(j>high)
while(i<=one_third)
{
u.at(k)=s.at(i);
i++;
k++;
}
if(i>one_third)
while(j<=high)
{
u.at(k)=s.at(j);
j++;
k++;
}
for(int n=low;n<k;n++)
s.at(n)=u.at(n);
}
}//end if
void mergesort(int low,int high){
if(low<high){
int one_third=(high-low)/3;//division,it's 3-way mergesort so obviously it's divided by 3
int two_third=2*one_third;
mergesort(low,one_third);
mergesort(one_third+1,two_third);
mergesort(two_third+1,high);
merge(low,one_third,two_third);
merge(low,two_third,high);
}//end if
}
at this point I guess I'm done thinking,Any answer/idea would be appreciated.

Answer 1

Here's a partial inspection of your code. I believe there is an issue debugging a 3 way merge sort with 4 values. You should use more values, such as 6 or 7.
Spaces not tabs for StackOverflow
I'll take a guess that the indentation is because you use tab characters in your code and pasted directly. You'll want to expand the tabs in your next post.
Precompiled Headers
Is your project huge? Does it significantly reduce the build time when you change a header or modify the source code?
I find that stdafx usually is more of a hassle and the time spent resolve defects it causes negates any potential savings by having a precompiled header.
Function prototypes should use named parameters
Can you tell the purpose of the different parameters in your declaration of merge and mergeSort?
Ambiguity breeds defects. 'nuff said.
Main function declared wrong.
The main function always returns an int to the operating system, always. The OS can ignore it.
This mechanism is so that script files can execute your program and test for errors.
Readability prevents defects
Invest in spaces around operators. The time saved by sacrificing spaces is negligible. The debugging time saved by having easy to read code is tremendous, especially when having other people review or inspect your code.
Use intermediate variables
Intermediate variables help clarify your program. They don't cost memory when you tell the compiler to optimize. During debugging, they can help show values during calculations.
The typical idiom for reading into a vector is:
int value;
cin >> value;
s.push_back(value);
The at method may have an overflow issue (or at least your not checking for out of bounds issues). The push_back method will cause the vector to expand as necessary.
Meaningful variable names reduces defects
The variable s has no meaning. Something like original_values or number_container are more descriptive. And again, variable name lengths have nothing to do with improving performance. Readable names help reduce the defects injected.
Not checking state of cin
If I enter "Lion" in response to your 2nd prompt, what will be in the 2nd slot of the array?
Don't trust the Users, they aren't perfect.
Don't clear the screen
It may contain useful data, such as the actual numbers entered. So when you are debugging, and want to know what the User actually typed in, it will be lost and gone forever.
Why cin.get twice?
You are asking the User for input without prompting. And twice. Bad Karma between your program and the User.
See cin.ignore if you want to ignore characters until a specific one is received. Something like this perhaps:
cout << "Paused. Press Enter to continue.\n";
cin.ignore(100000, '\n');
Magic numbers
In function mergesort, you use the numbers 2 and 3. Why? What's their purpose?
Redundant comments
Most programmers realize that the '/' character in a math expression is division. The comment is redundant.
Also, why divide by 3? It's a nasty number. Do you realize you are performing integer division and your product will be truncated? For example: 1/3 == 2/3 == 0.
USE A DEBUGGER
Lastly, a lot of your program's functionality can be verified easier and quicker by using a debugger. A debugger allows you to execute a statement and see the variable values. You can set breakpoints to stop execution at different places. It's a worthwhile educational investment, start now.

Answer 2

A "classic" 3 way merge sort merges runs 3 at a time, alternating between a source and destination array (or vector or list). The code needs to perform up to 3 compares in order to determine the "smallest" of 3 values from each of the 3 runs, then move the smallest value from it's corresponding run to the destination array. The code also has to handle the case where the end of a run is reached, leaving only 2 runs to merge, then the case where the end of the second run is reached, in which case the rest of the third run is moved to the destination array.
For a ram based sort, I'm not sure this is any faster than a normal 2 way merge. For an external sort, with multiple devices or very large read and writes, then a k way merge with k up to 12 or 16 will be faster.