Stimmt. Das habe ich nicht bedacht. Allerdings sind bei mir die Zahlen für die Modulo-Variante noch immer besser (wenn auch nicht mehr so drastisch):
Code: Alles auswählen
#!/usr/bin/env python3
from time import time
from random import randint, shuffle
from statistics import mean, stdev
N = 10_000
REPETITIONS = 100
LOWER = 1
UPPER = 1_000_000_000
def string_based(s):
return s[-1] in ('0', '5') and len(s) > 2
def modulo_based(n):
return n % 5 == 0 and n > 99
def decimal_based(n):
if n % 10 not in (0, 5):
return False
i = 0
k = n
while k > 0:
i += 1
k //= 10
if i > 2:
return False
return True
ALGORITHMS = [
('string comparison', string_based),
('modulo calculation', modulo_based),
('decimal position', decimal_based)
]
def random_numbers(n, lower, upper):
divisible = set()
indivisible = set()
while len(divisible) < n or len(indivisible) < n:
x = randint(lower, upper)
if len(divisible) < n and x % 5 in (0, 5):
divisible.add(x)
elif len(indivisible) < n:
indivisible.add(x)
return tuple(divisible), tuple(indivisible)
def main():
print('N:', N)
print('Range:', LOWER, UPPER)
print('Repetitions:', REPETITIONS)
results = dict()
algorithms = ALGORITHMS
for _ in range(REPETITIONS):
divisible, indivisible = random_numbers(N, LOWER, UPPER)
for label, function in ALGORITHMS:
if label == 'string comparison':
divisible_temp = [str(i) for i in divisible]
indivisible_temp = [str(i) for i in indivisible]
else:
divisible_temp = divisible
indivisible_temp = indivisible
t1 = time()
for group in (divisible_temp, indivisible_temp):
for number in group:
result = function(number)
t2 = time()
results.setdefault(label, [])
results[label].append(t2 - t1)
# shuffle last, so order is mainted in modern python
shuffle(algorithms)
print()
for label, durations in results.items():
print(label, '(mean):', mean(durations))
print(label, '(sd):', stdev(durations))
print(label, '(sum):', sum(durations))
print()
if __name__ == '__main__':
main()
Code: Alles auswählen
$ ./fives.py :(
N: 10000
Range: 1 1000000000
Repetitions: 100
string comparison (mean): 0.0033177709579467773
string comparison (sd): 0.00032582172245825377
string comparison (sum): 0.33177709579467773
modulo calculation (mean): 0.002630734443664551
modulo calculation (sd): 0.0005114850119894751
modulo calculation (sum): 0.2630734443664551
decimal position (mean): 0.0063778114318847655
decimal position (sd): 0.0006249521611702232
decimal position (sum): 0.6377811431884766
Das Result meiner letzten Messung ist ja nun, dass es besser ist, Zahlen als Zahlen vorliegen zu haben. Also kann man festhalten, Modulo ist für Zahlen absolut gesehen schneller als die String-Variante für Strings. Allerdings ist das eigentliche Bottleneck die Konvertierung und der Fragsteller sollte die String-Variante nehmen.
