Counting inversion using merge sort

Question

I know that there was a plenty of these implementation here in Stack, but I have got a problem which I cannot handle.

First I have implemented the merge sort at khanacademy with javascript, then I rewrite the code to C++, and tried to count number of inversion in an array.

I did the best I could, and spent an hour trying to get understand what have I done wrong. I did search for another implementation here at stack and tried to correct my code. Unfortunately I do not know what am I doing wrong.I think that I count every inversion. Thanks in advance for an assistance with having understood what is wrong.

My code:

int lowhalflength(int p, int q)
{
    return q - p + 1;
}

int highhalflength(int q, int r)
{
    return r - q;
}


int merge(int array[], int p, int q, int r, int lowhalf[], int highhalf[])
{
    int k = p;
    int i;
    int j;
    int count = 0;
    for (int i = 0; k <= q; i++ , k++)
    {
        lowhalf[i] = array[k];
    }
    for (int i = 0; k <= r; i++ , k++)
    {
        highhalf[i] = array[k];
    }

    k = p;
    i = 0;
    j = 0;

    while (i <= (q - p) && j <= r - (q + 1))
    {
        if (lowhalf[i] <= highhalf[j])
        {
            array[k] = lowhalf[i];
            i++;
        }
        else
        {
            array[k] = highhalf[j];
            j++;
            count += q - 1;
        }

        k++;
    }

    while (i < lowhalflength(p, q))
    {
        array[k] = lowhalf[i];
        k++;
        i++;
    }

    while (j < highhalflength(q, r))
    {
        array[k] = highhalf[j];
        k++;
        j++;
    }


    return count;
}

The mergeSort function:

int mergeSort(int array[], int p, int r)
{
    int q = ((p + r) / 2);
    int* lowhalf = new int[lowhalflength(p, q)];
    int* highhalf = new int[highhalflength(q, r)];

    int count = 0;
    if (p < r)
    {
        q = ((p + r) / 2);
        count = mergeSort(array, p, q);
        count += mergeSort(array, q + 1, r);
        count += merge(array, p, q, r, lowhalf, highhalf);
    }
    delete[] lowhalf;
    delete[] highhalf;
    return count;
}

For array [10, 9, 8, 7, 6, 5, 4, 3, 2, 1] the output is 46 while it should be 45.

EDIT: The answer is to change the following line q-1 to q+jk. I found it by myself but I do not know how should I interpret it. Any hint or proof why it is correct will be really desirable.

Answer 1

You can use my code for counting inversion pair and your merge function should look like this in more efficient manner as:

int merge(int *array, int lower, int mid, int upper) {

    // Initialisation of the sizes of two subarrays and subarrays also.
    int left_array_size = mid - lower + 1;
    int right_array_size = upper - mid;
    int left_array[left_array_size], right_array[right_array_size];

    int j = 0;
    for (int i = lower; i <= mid; i++) {
        left_array[j++] = array[i];
    }
    j = 0;
    for (int i = mid + 1; i <= upper; i++) {
        right_array[j++] = array[i];
    }

    // Performing merging in a non-increasing manner and count inversion pairs..
    int i = 0, k;
    j = 0;
    int resultIntermediate = 0;
    for (k = lower; k <= upper; ) {
        if (left_array[i] <= right_array[j]) {
            array[k++] = left_array[i++];
            if (i >= left_array_size)   break;
        }
        else {
            array[k++] = right_array[j++];

            // If a element in left_array_size is greater than an element from
            // right_array_size then rest of all other elements will also be
            // greater than that element of right_array_size because both
            // subarrays are sorted in non-decreasing order.
            resultIntermediate += left_array_size - i;

            if (j >= right_array_size)  break;
        }
    } //end of for loop.


    // Performing merging if i or j doesn't reach to its
    // maximum value i.e. size of the subarrays.
    while (i < left_array_size) {
        array[k++] = left_array[i++];
    }
    while (j < right_array_size) {
        array[k++] = right_array[j++];
    }

    // Returning the result...
    return resultIntermediate;

} //end of the merge function.

And function to count inversion pair

int countInversionPair(int *array, int lower, int upper) {
    int count_inv_pair = 0;
    // Do recusion untill the problem / array can be subdevided.
    if (lower < upper) {

        // Partition the Array into two subproblems.
        int mid = (lower + upper) / 2;

        // Call the countInversionPair() function for these two
        // subarrays / subproblems recursively to count number of
        // inversion for these subproblems / subarrays.
        count_inv_pair = countInversionPair(array, lower, mid);
        count_inv_pair += countInversionPair(array, mid + 1, upper);

        // Merge these two subarrays into a sigle array
        count_inv_pair += merge(array, lower, mid, upper);
    }
    return count_inv_pair;
}

Now you can get number of inversion pair by calling the function from main as:

int count_inv_pair = countInversionPair(array, 0, size - 1);

And now you will get your answer..

Answer 2

Big thanks for all of you guys, especially @Shiv and @WhozCraig, you gave me and idea how to solve it. The answer is to change q-1 to q+jk