How to Implement a Stack in Python and Check the Run-time Performance

We will cover the following in this article

  • What is a Stack – a short introduction
  • How to implement a Stack in Python
  • Investigate the run-time performance

What is a Stack

A Stack is a useful concept that is used in daily life, and hence, a concept that is important to understand and master as a programmer.

To understand Stacks just think of a stack of plates. There are two main operations you can do. First, you can add a plate on top of the stack. That operation is called push adds the element on top of the stack. Second, you can remove the top plate of the stack. That operation is called pop, and returns the top element of the stack.

In the diagram below a Stack is pictured. It contains of a Stack of element on the left side. The operation push of the element 03 is executed and results is pictured on the right side. Notice, that the push operation puts the element on top of the stack.

Below the operation pop is executed. Notice, that the pop operation takes from the top of the stack.

Implementation of a Stack in Python

It is a good idea to have a helper class Node that represents the elements on the stack.

class Node:
    def __init__(self, element=None, next_node=None):
        self.element = element
        self.next_node = next_node

The actual functionality of the Stack is kept in a Stack class.

class Stack:
    def __init__(self):
        self.stack = None

    def push(self, element):
        self.stack = Node(element, self.stack)

    def pop(self):
        element = self.stack.element
        self.stack = self.stack.next_node
        return element

    def is_empty(self):
        return self.stack is None

Now you can use your stack. Like the example below.

s = Stack()
for i in range(5):
while not s.is_empty():

Will give the following output.


Notice the order of the element being removed from the stack by pop.

Run-time Performance

If we look at how the stack perform in order of the data size. To investigate the run-time performance the cProfile library is a good choice and simple to use. The following piece of code will help you investigate the performance.

import cProfile

def profile_stack(n):
    s = Stack()
    for i in range(n):
    while not s.is_empty():

See the following graph.

As you see, the push and pop operations are constant, O(1), resulting in a linear performance of n push and pop operations as in the above experiment.

Comparing Performance of Python list as a Stack – How a wrong implementation can ruin performance

A Stack?

A Stack is using the principle first-in-last-out.

It is like a stack of plates. The last one you put on the top is the first one you take.

How can you implement them in Python? Well, we are in luck, you can use a Stack, and if done correctly, you will have the same performance as an actual Stack implementation will have.

But first, how can you do it wrong?

Well, you might think that the first element of the list is the top of your stack, hence in you will insert the elements on the first position, and, hence, remove them from the first position as well.

# Create a list as a stack
s = []

# Insert into the first position.
element = 7
s.insert(0, element)

# Remove from the first position.

Sounds about right?

Let’s test that and compare it with a different approach. To add the newest element to the end of the list, and, hence, remove them from the end of the list.

# Create a list and use it as stack
s = []

# Insert element in last postion
element = 7

# Remove from the last position

Let’s check the performance of those two approaches.

Comparing the performance of the two approaches

How do you compare. You can use cProfile library. It is easy to use and informative results

See the sample code below, which compares the two approaches by create a stack each and inserting n elements to it and removing them afterwards.

import cProfile

def profile_list_as_queue_wrong(n):
    s = []
    for i in range(n):
        s.insert(0, i)
    while len(s) > 0:

def profile_list_as_queue_correct(n):
    s = []
    for i in range(n):
    while len(s) > 0:

def profile(n):

The results are given here.

   Ordered by: standard name

   ncalls  tottime  percall  cumtime  percall filename:lineno(function)
        1    0.000    0.000    5.842    5.842 <string>:1(<module>)
        1    0.078    0.078    0.107    0.107
        1    0.000    0.000    5.842    5.842
        1    0.225    0.225    5.735    5.735
   200002    0.017    0.000    0.017    0.000 {len}
   100000    0.007    0.000    0.007    0.000 {method 'append' of 'list' objects}
        1    0.000    0.000    0.000    0.000 {method 'disable' of '_lsprof.Profiler' objects}
   100000    3.547    0.000    3.547    0.000 {method 'insert' of 'list' objects}
   200000    1.954    0.000    1.954    0.000 {method 'pop' of 'list' objects}
        2    0.014    0.007    0.014    0.007 {range}

Observe that the “wrong” implementation takes over 5 seconds and the “correct” takes approximately 0.1 second. Over a factor 50 in difference.

Looking into the details

If we look at the complexities given by Python, it explains it all.

The Python lists amortised complexities are given on this page.

And you notice that the append and pop (last element) are O(1), which means constant time. Constant time, means that the operations are independent on the size of the lists. That means the correct implementation gives O(n) time complexity.

On the other hand, the insert and pop(0) have linear performance. That basically means that we with the wrong implementation end up with O(n^2) time complexity.