A simple experiment in Python comparing performance of two approaches to solve the same problem.

In the following examples, we will have a list named numbers containing integers from 0 to 4, and our task will be to remove all occurrences of 4 from the list in-place.

This will be the LBYL approach:

while 4 in numbers:
    numbers.remove(4)

And this will be the EAFP approach:

while True:
    try:
        numbers.remove(4)
    except ValueError:
        break

In theory, the former approach should be less efficient compared to the latter because with each iteration, the loop condition will perform an additional item look-up in the list. Let’s verify this assumption.

Processing a short list

Scenario:

  • list length: 5,000 items
  • number of repetitions: 100,000

The timeit.timeit function will be used to perform the operation many times repeatedly and measure the execution time.

from timeit import timeit

setup = 'numbers = list(range(5)) * 1000'

lbyl_code = '''
while 4 in numbers:
    numbers.remove(4)
'''

eafp_code = '''
while True:
    try:
        numbers.remove(4)
    except ValueError:
        break
'''

lbyl_time = timeit(lbyl_code, setup=setup, number=100_000)
eafp_time = timeit(eafp_code, setup=setup, number=100_000)

print(f'LBYL: {lbyl_time:.3f} s')
print(f'EAFP: {eafp_time:.3f} s')

Output of three sample runs of the program:

LBYL: 6.232 s
EAFP: 5.892 s

LBYL: 6.221 s
EAFP: 5.953 s

LBYL: 6.233 s
EAFP: 5.903 s

We can see that the try-except approach is indeed faster, but the difference is not significant.

Processing a long list

Scenario:

  • list length: 50,000 items
  • number of repetitions: 1

The codetiming third-party library will be used to perform the operation once without repetition and measure the execution time.

from codetiming import Timer

numbers = list(range(5)) * 10_000

with Timer(text='LBYL: {:.3f} s'):
    while 4 in numbers:
        numbers.remove(4)


numbers = list(range(5)) * 10_000

with Timer(text='EAFP: {:.3f} s'):
    while True:
        try:
            numbers.remove(4)
        except ValueError:
            break

Output of three sample runs of the program:

LBYL: 6.104 s
EAFP: 2.999 s

LBYL: 6.334 s
EAFP: 3.010 s

LBYL: 6.099 s
EAFP: 3.008 s

We can see that the difference in execution time is significant in this case: the try-except approach is twice as faster.

Final note

The LBYL solution in this simple example is less efficient because of the additional 4 in numbers list item look-up. This operation’s average time complexity is O(n).