A Simplified Explanation of Time and Space Complexity of Merge Sort

Updated on June 27, 2024

Detailed Breakdown of Merge Sort’s Time Complexity

Best Case Time Complexity

In ideal circumstances, the array is already arranged. Merge Sort keeps working even if it seems like we won’t need to sort it again. Since the elements are already sorted, fewer comparisons are needed when the array is divided and merged as usual.

Let’s consider an already sorted array: [1, 2, 3, 4, 5, 6, 7].

• Divide Phase:
  • [1, 2, 3]
  • [4, 5, 6, 7]
• Conquer Phase:
  • [1, 2, 3] becomes [1], [2], [3]
  • [4, 5, 6, 7] becomes [4], [5], [6], [7]
• Combine Phase:
  • [1] and [2] merge to [1, 2], then [1, 2] and [3] merge to [1, 2, 3]
  • [4] and [5] merge to [4, 5], then [4, 5] and [6] merge to [4, 5, 6], then [4, 5, 6] and [7] merge to [4, 5, 6, 7]
  • Finally, [1, 2, 3] and [4, 5, 6, 7] merge to [1, 2, 3, 4, 5, 6, 7]

Because the elements are already ordered, fewer comparisons are made during the merging process in this case. However, due to the splitting and merging processes, the time complexity remains O(n log n).

Average Case Time Complexity

The most typical case occurs when the components are jumbled together but not completely out of order. Here, the comparisons and swaps will be distributed more evenly throughout the operation.

Consider an array: [10, 5, 2, 8, 7, 3, 6].

• Divide Phase:
  • [10, 5, 2]
  • [8, 7, 3, 6]
• Conquer Phase:
  • [10, 5, 2] becomes [10], [5], [2]
  • [8, 7, 3, 6] becomes [8], [7], [3], [6]
• Combine Phase:
  • [10] and [5] merge to [5, 10], then [5, 10] and [2] merge to [2, 5, 10]
  • [8] and [7] merge to [7, 8], then [7, 8] and [3] merge to [3, 7, 8], then [3, 7, 8] and [6] merge to [3, 6, 7, 8]
  • Finally, [2, 5, 10] and [3, 6, 7, 8] merge to [2, 3, 5, 6, 7, 8, 10]

The time complexity is still O(n log n) in this typical situation. Splitting and merging are done in each phase, along with an equal number of swaps and comparisons.

Worst Case Time Complexity

When the array is sorted in reverse order, the worst-case situation occurs. In this scenario, the quantity of comparisons and swaps required during the merging processes is maximised.

Consider an array: [7, 6, 5, 4, 3, 2, 1].

• Divide Phase:
  • [7, 6, 5]
  • [4, 3, 2, 1]
• Conquer Phase:
  • [7, 6, 5] becomes [7], [6], [5]
  • [4, 3, 2, 1] becomes [4], [3], [2], [1]
• Combine Phase:
  • [7] and [6] merge to [6, 7], then [6, 7] and [5] merge to [5, 6, 7]
  • [4] and [3] merge to [3, 4], then [3, 4] and [2] merge to [2, 3, 4], then [2, 3, 4] and [1] merge to [1, 2, 3, 4]
  • Finally, [5, 6, 7] and [1, 2, 3, 4] merge to [1, 2, 3, 4, 5, 6, 7]

Even in this worst case, the time complexity remains O(n log n). The key difference lies in the number of comparisons and swaps required to merge the sub-arrays.

In-depth Analysis of Merge Sort’s Space Complexity

General Space Complexity

Merge Sort is known for its efficient time complexity, but it also requires additional memory. Its space complexity is O(n). This means that, in addition to the input array, we need extra space equal to its size.

Why does Merge Sort need extra space? When we merge two sorted halves, we use an auxiliary array to hold the merged result before copying it back to the original array. This auxiliary array is crucial for the merge process. Let’s look at an example to understand this better.

Example of Space Usage

Consider an array: [38, 27, 43, 3, 9, 82, 10].

1. Divide Phase:
   • [38, 27, 43]
   • [3, 9, 82, 10]
2. Conquer Phase:
   • [38, 27, 43] becomes [38], [27], [43]
   • [3, 9, 82, 10] becomes [3], [9], [82], [10]
3. Combine Phase:
   • [38] and [27] merge to [27, 38], then [27, 38] and [43] merge to [27, 38, 43]
   • [3] and [9] merge to [3, 9], then [3, 9] and [82] merge to [3, 9, 82], then [3, 9, 82] and [10] merge to [3, 9, 10, 82]
   • Finally, [27, 38, 43] and [3, 9, 10, 82] merge to [3, 9, 10, 27, 38, 43, 82]

Throughout the merge process, we use extra space equivalent to the size of the array.

Using Linked Lists to Optimise Space

Can we optimise space complexity? Yes, using linked lists can help. With linked lists, we can merge in place, avoiding the need for a large auxiliary array. Instead, we only need space for the recursion stack. This reduces the space complexity to O(log n).

Using the same array but represented as a linked list, the merge process works directly on the list nodes, reducing space overhead. This approach is beneficial in memory-constrained environments. However, implementing Merge Sort with linked lists can be more complex.

Recursion and Stack Frames

Merge Sort is a recursive algorithm. Each recursive call adds a frame to the stack, contributing to space complexity. The maximum depth of recursion is log n, so the stack space needed is O(log n). While this is less than the O(n) for the auxiliary array, it still impacts the overall space complexity.

Practical Considerations

In practical applications, the space complexity of Merge Sort affects performance, especially with large datasets. The extra space requirement can be a drawback compared to in-place sorting algorithms like Quick Sort. However, the stability and predictable performance of Merge Sort often outweigh these considerations.

Practical Implications of Parallelising Merge Sort

Parallelising Merge Sort can potentially speed up the sorting process. By dividing the work among multiple processors, we can sort large datasets more efficiently. But what are the real benefits and challenges of this approach?

Benefits of Parallelising Merge Sort

Parallelising Merge Sort can significantly reduce sorting time. Each processor handles a part of the array independently. After sorting their parts, processors then merge the results. This parallel processing can be faster, especially for large datasets.

Challenges of Parallelising Merge Sort

While the theory sounds promising, there are practical challenges. Synchronisation between processors can introduce overhead. Managing multiple processors also adds complexity. Moreover, not all systems support efficient parallel processing.

Example of Parallel Merge Sort

Let’s look at a simple example using Python:

def merge_sort_parallel(arr, pool):

if len(arr) > 1:

mid = len(arr) // 2

left_half = arr[:mid]

right_half = arr[mid:]

pool.apply_async(merge_sort_parallel, (left_half, pool))

pool.apply_async(merge_sort_parallel, (right_half, pool))

left_half = merge_sort_parallel(left_half, pool)

right_half = merge_sort_parallel(right_half, pool)

i = j = k = 0

while i < len(left_half) and j < len(right_half):

if left_half[i] < right_half[j]:

arr[k] = left_half[i]

i += 1

else:

arr[k] = right_half[j]

j += 1

k += 1

while i < len(left_half):

arr[k] = left_half[i]

i += 1

k += 1

while j < len(right_half):

arr[k] = right_half[j]

j += 1

k += 1

return arr

if __name__ == “__main__”:

arr = [38, 27, 43, 3, 9, 82, 10]

pool = multiprocessing.Pool(processes=4)

sorted_arr = merge_sort_parallel(arr, pool)

pool.close()

pool.join()

print(sorted_arr)

Output

The output will be:

Practical Considerations

Before deciding to parallelise Merge Sort, consider the hardware and the size of the dataset. For small arrays, the overhead may not justify the speedup. For larger datasets on suitable hardware, the benefits can be substantial.

Comparison of Merge Sort with Other Sorting Algorithms

Quick Sort vs. Merge Sort

Both Quick Sort and Merge Sort are very efficient data sorting algorithms despite the fact that they have different capabilities. Because the overhead is kept to a minimum, Quick Sort is relatively faster. But, one should also keep in mind that it has a worst-case time complexity of O(n²) while Merge Sort has O(n log n). One more difference that needs to be highlighted is that while Merge Sort is an example of a stable algorithm, meaning that it keeps the order of equal elements intact, Quick Sort is an example of an unstable sorting algorithm.

Insertion Sort vs. Merge Sort

The insertion sort algorithm is best suited for small numbers of elements and partially sorted elements. In the best case, it has a time complexity of O(n). Thus, Merge Sort is preferable for larger data collections due to its O(n log n) time complexity. Like many sorting algorithms, Insertion Sort is often used in conjunction with other algorithms for small subarrays.

Bubble Sort vs. Merge Sort

Bubble Sort is a data sorting algorithm which is very easy to understand and implement, but because of its time complexity O(n²), efficiency is very low. This is the reason why it can’t be used for larger data sets. Therefore, for large arrays, Merge Sort, with its time complexity of O(n log n), is more efficient. Bubble sort is useful for educational settings and small arrays since its technique is very simple to implement.

Comparison Table