consider-math-not-float / R6106

Message emitted:

Consider %smath.%s instead of %s

Description:

Using math.inf or math.nan permits to benefit from typing and it is up to 4 times faster than a float call (after the initial import of math). This check also catches typos in float calls as a side effect.

Problematic code:

swag = float("inf")  # [consider-math-not-float]

Correct code:

import math

swag = math.inf

Configuration file:

[MAIN]
load-plugins=pylint.extensions.code_style

Additional details:

This is an extension check because the typing advantage could be fixed.

Regarding performance, float("nan") and float("inf") are slower than their counterpart math.inf and math.nan by a factor of 4 after the initial import of math.

import math
import timeit

time_math_inf = timeit.timeit('math.nan', globals=globals(), number=10**8)
print(f'math.nan: {time_math_inf:.2f} seconds')

import timeit
time_inf_str = timeit.timeit('float("nan")', number=10**8)
print(f'float("nan"): {time_inf_str:.2f} seconds')

Result:

math.nan: 1.24 seconds
float("nan"): 5.15 seconds

But if we take the initial import into account it's worse.

import timeit

time_math_inf = timeit.timeit('import math;math.nan', globals=globals(), number=10**8)
print(f'math.nan: {time_math_inf:.2f} seconds')

import timeit
time_inf_str = timeit.timeit('float("nan")', number=10**8)
print(f'float("nan"): {time_inf_str:.2f} seconds')

Result:

math.nan: 9.08 seconds
float("nan"): 5.33 seconds

So the decision depends on how and how often you need to use it and what matter to you.

Note

This message is emitted by the optional 'code_style' checker, which requires the pylint.extensions.code_style plugin to be loaded.

Created by the code_style checker.