Performance

Bubble sort has a worst-case and average complexity of ${\displaystyle O(n^{2})}$, where ${\displaystyle n}$ is the number of items being sorted. Most practical sorting algorithms have substantially better worst-case or average complexity, often ${\displaystyle O(n\log n)}$. Even other ${\displaystyle O(n^{2})}$ sorting algorithms, such as insertion sort, generally run faster than bubble sort, and are no more complex. For this reason, bubble sort is rarely used in practice.

Like insertion sort, bubble sort is adaptive, giving it an advantage over algorithms like quicksort. This means that it may outperform those algorithms in cases where the list is already mostly sorted (having a small number of inversions), despite the fact that it has worse average-case time complexity. For example, bubble sort is ${\displaystyle O(n)}$ on a list that is already sorted, while quicksort would still perform its entire ${\displaystyle O(n\log n)}$ sorting process.

While any sorting algorithm can be made ${\displaystyle O(n)}$ on a presorted list simply by checking the list before the algorithm runs, improved performance on almost-sorted lists is harder to replicate.

Rabbits and turtles

The distance and direction that elements must move during the sort determine bubble sort's performance because elements move in different directions at different speeds. An element that must move toward the end of the list can move quickly because it can take part in successive swaps. For example, the largest element in the list will win every swap, so it moves to its sorted position on the first pass even if it starts near the beginning. On the other hand, an element that must move toward the beginning of the list cannot move faster than one step per pass, so elements move toward the beginning very slowly. If the smallest element is at the end of the list, it will take ${\displaystyle n-1}$ passes to move it to the beginning. This has led to these types of elements being named rabbits and turtles, respectively, after the characters in Aesop's fable of The Tortoise and the Hare.

Various efforts have been made to eliminate turtles to improve upon the speed of bubble sort. Cocktail sort is a bi-directional bubble sort that goes from beginning to end, and then reverses itself, going end to beginning. It can move turtles fairly well, but it retains ${\displaystyle O(n^{2})}$ worst-case complexity. Comb sort compares elements separated by large gaps, and can move turtles extremely quickly before proceeding to smaller and smaller gaps to smooth out the list. Its average speed is comparable to faster algorithms like quicksort.

Step-by-step example

Take an array of numbers "5 1 4 2 8", and sort the array from lowest number to greatest number using bubble sort. In each step, elements written in bold are being compared. Three passes will be required;

First Pass

( 5 1 4 2 8 ) → ( 1 5 4 2 8 ), Here, algorithm compares the first two elements, and swaps since 5 > 1.

( 1 5 4 2 8 ) → ( 1 4 5 2 8 ), Swap since 5 > 4

( 1 4 5 2 8 ) → ( 1 4 2 5 8 ), Swap since 5 > 2

( 1 4 2 5 8 ) → ( 1 4 2 5 8 ), Now, since these elements are already in order (8 > 5), algorithm does not swap them.

Second Pass

( 1 4 2 5 8 ) → ( 1 4 2 5 8 )

( 1 4 2 5 8 ) → ( 1 2 4 5 8 ), Swap since 4 > 2

( 1 2 4 5 8 ) → ( 1 2 4 5 8 )

( 1 2 4 5 8 ) → ( 1 2 4 5 8 )

Now, the array is already sorted, but the algorithm does not know if it is completed. The algorithm needs one additional whole pass without any swap to know it is sorted.

Third Pass

( 1 2 4 5 8 ) → ( 1 2 4 5 8 )

( 1 2 4 5 8 ) → ( 1 2 4 5 8 )

( 1 2 4 5 8 ) → ( 1 2 4 5 8 )

( 1 2 4 5 8 ) → ( 1 2 4 5 8 )

Pseudocode implementation

In pseudocode the algorithm can be expressed as (0-based array):

procedure bubbleSort(A : list of sortable items)
    n := length(A)
    repeat
        swapped := false
        for i := 1 to n-1 inclusive do
            { if this pair is out of order }
            if A[i-1] > A[i] then
                { swap them and remember something changed }
                swap(A[i-1], A[i])
                swapped := true
            end if
        end for
    until not swapped
end procedure

Optimizing bubble sort

The bubble sort algorithm can be optimized by observing that the n-th pass finds the n-th largest element and puts it into its final place. So, the inner loop can avoid looking at the last n − 1 items when running for the n-th time:

procedure bubbleSort(A : list of sortable items)
    n := length(A)
    repeat
        swapped := false
        for i := 1 to n - 1 inclusive do
            if A[i - 1] > A[i] then
                swap(A[i - 1], A[i])
                swapped := true
            end if
        end for
        n := n - 1
    until not swapped
end procedure

More generally, it can happen that more than one element is placed in their final position on a single pass. In particular, after every pass, all elements after the last swap are sorted, and do not need to be checked again. This allows us to skip over many elements, resulting in about a worst case 50% improvement in comparison count (though no improvement in swap counts), and adds very little complexity because the new code subsumes the "swapped" variable:

To accomplish this in pseudocode, the following can be written:

procedure bubbleSort(A : list of sortable items)
    n := length(A)
    repeat
        newn := 0
        for i := 1 to n - 1 inclusive do
            if A[i - 1] > A[i] then
                swap(A[i - 1], A[i])
                newn := i
            end if
        end for
        n := newn
    until n ≤ 1
end procedure

Alternate modifications, such as the cocktail shaker sort attempt to improve on the bubble sort performance while keeping the same idea of repeatedly comparing and swapping adjacent items.