Improve performance of expo by multiplying previous result

[whonore]

This computes the exponential by multiplying the previous result by base instead of computing it from scratch with ** each time, which seems to be about an order of magnitude faster. On my machine the following tests report around 2.63 seconds for expo_old and 0.10 for expo_new.

def expo_old(base=2, factor=1, max_value=None):
    n = 0
    while True:
        a = factor * base ** n
        if max_value is None or a < max_value:
            yield a
            n += 1
        else:
            yield max_value

def expo_new(base=2, factor=1, max_value=None):
    a = 1
    while True:
        if max_value is None or a < max_value:
            yield factor * a
            a *= base
        else:
            yield max_value

import timeit

for e in (expo_old(), expo_new()):
    print(timeit.timeit('for i in range(50): next(e)', globals={'e': e}, number=1000))

[bgreen-litl]

This seems good to me.

Just curious, did you run into this because you in your use case the number of retries was very high?

[whonore]

No, I was just browsing the code and noticed that some work was unnecessarily re-done on each loop. I think it's pretty unlikely that anyone would reach enough retries for this to actually make much of a difference, but I figured I'd submit it anyway since it's such a simple change.

[gaby]

Any chance this can get merged? PR has been open for over a year.

[whonore]

Ping?

[frost-nzcr4]

This max_value condition doesn't work properly with your changes, it needs to be updated a bit, like this, but still 24x faster than the original:

    base_n = 1
    while True:
        a = factor * base_n
        if max_value is None or a < max_value:
            yield a
            base_n *= base

[whonore]

Good catch, thanks.