Repository navigation
minor inefficency in the statistics module #154947
Description
Bug report
Bug description:
Line 1217 in 22a6c51
in the quantiles computation 'i * m' should be reused not recomputed ( #150326) changes not added here)
i_m = i * m
j = i_m // n # rescale i to m/n
j = 1 if j < 1 else ld-1 if j > ld-1 else j # clamp to 1 .. ld-1
delta = i_m - j*n
CPython versions tested on:
3.13
Operating systems tested on:
No response
Activity
picnixz commentedon Jul 30, 2026
If you believe there is a real improvement this should be backed by benchmarks. I do not think this matters though.
lbaumbauer commentedon Jul 30, 2026
lol
`from timeit import timeit
a = 1.23456789
def test_a(ld, i, m):
n = 100
i_m = i * m
j = i_m // n # rescale i to m/n
j = 1 if j < 1 else ld-1 if j > ld-1 else j # clamp to 1 .. ld-1
delta = i_m - j*n
def test_b(ld, i, m):
n = 100
j = i * m // n # rescale i to m/n
j = 1 if j < 1 else ld-1 if j > ld-1 else j # clamp to 1 .. ld-1
delta = i * m - j*n
_ld = 234
_i = 3
_m = _ld + 1
x = timeit(lambda : test_a(_ld, _i, _m), number=10_000_000)
y = timeit(lambda : test_b(_ld, _i, _m), number=10_000_000)
print(f"{x / y}")
`
0.925picnixz commentedon Jul 31, 2026
This benchmark is neither correct nor relevant; I am closing this. Use pyperf for more robust checks and benchmark the entire function as well instead of just thjs arithmetic part. Timeit does not givr any information (and use PGO/LTO builds).