This avoids unnecessary operations, such as clone, multiplication, and division.

For the following microbenchmark, the current master would take time 3.41s, 0.73s, 3.69s, and it takes 2.27s, 0.50s, 2.60s after applying this patch.

import time

t0 = time.time()
for i in range(10000000):
    pass
t1 = time.time()
print(t1 - t0)

t0 = time.time()
for i in range(0, 10000000, 5):
    pass
t1 = time.time()
print(t1 - t0)

t0 = time.time()
for i in range(10000000, 0, -1):
    pass
t1 = time.time()
print(t1 - t0)

Previously, the PyRange::get function takes 30% of the time, and it takes only 10% of execution time after this optimization.

Currently, 34% of the time is spent on PyDict::inner_itemset_fast (a lot of time spent on dropping the old value), and about 10% is spent on BigInt conversion from usize and to PyInt.