Don't use the walrus operator := in convoluted ways!

import sys

if (i := input())[0] == "q" or i == "exit":
    sys.exit()

This snippet of code stops the program if the input starts with a "q" or if the input is "exit". But does it look good?

The walrus operator is, at the time of writing, a fairly recent addition to Python (you can read about its introduction to the language in the PEP 572). This means people are still getting used to it and haven't really grasped how to use it well. Again, PEP 572 includes a really nice explanation of why the walrus operator was introduced in Python and provides a fair share of really good use cases.

The pythonic version of the code above would be

import sys

i = input()
if i[0] == "q" or i == "exit":
    sys.exit()

It uses one more line but it does keep it much more readable. Not only readability counts but this is also a case where one can argue that beautiful is better than ugly!

Personally, I think the walrus operator := is particularly useful when we want to write a while loop where the body uses something that is also used in the loop condition, which also needs an initialization step. For example, say you want to write a simple loop that checks if the input line is not empty and, in that case, evaluate it as Python code. Without using := there are a couple of decent alternatives, like

inp = input()
while inp:
    eval(inp)
    inp = input()

or

while True:
    inp = input()
    if not inp:
        break
    eval(inp)

but none of those snippets beat this alternative implementation that is shorter, more expressive and much more elegant:

while (inp := input()):
    eval(inp)

Below you can find a couple of other examples of interesting use cases of :=, taken and/or adapted from PEP 572.

Look for a "witness" in a list

import random
random.seed(0)  # for reproducibility

ints = [random.randint(0, 100) for _ in range(100)]
if any((witness := elem) % 10 == 0 for elem in ints):
    print(f"{witness} is the first multiple of 10 in the list!")
else:
    print("No multiples of 10 found.")

This snippet prints 100 is the first multiple of 10 in the list!, because it goes through the randomly generated list ([49, 97, 53, 5, 33, 65, 62, 51, 100, 38, ...]) and searches for a multiple of 10. The first multiple found is 100.

Track changes in a list comprehension

import random
random.seed(0)  # for reproducibility

some_list = [random.randint(-100, 100) for _ in range(50)]
total = 0
partial_sums = [total := total + elem for elem in some_list]
print(partial_sums)

This snippets prints [-2, 92, 99, 9, -25, ...] which are the first \(5\) partial sums of some_list, whose values are [-2, 94, 7, -90, -34, ...].

Reuse potentially expensive computations

def fib(n):
    """This could be made more efficient in many different ways!"""
    if n <= 1:
        return n
    else:
        return n * fib(n-1)

n = 17
print(f"Fib {n} is {(f:=fib(n))} and fib {n+1} is {(n+1)*f}")

This snippet prints Fib 17 is 355687428096000 and fib 18 is 6402373705728000, corresponding to \(17!\) and \(18! = 18 \times 17!\).


If you find a situation where you really wanted to use the walrus operator but it doesn't really fit well with the surrounding code... Then maybe that is not the time nor the place! Don't disrespect the Zen of Python!

Did this make any sense? Let me know in the comment section below!

Previous Post

Add a comment