PSA: Python Float Overflow
Python 2 and Python 3 provide arbitrary precision integers. This means that you don’t need to worry about overflowing or underflowing the integer datatypes, ever.
2**10000? No problem. It makes it pretty convenient to write code. Floats, however, are not arbitrary precision:
>>> a = 2 ** 10000 >>> a * .1 Traceback (most recent call last): File "<stdin>", line 1, in <module> OverflowError: int too large to convert to float
infinity. It’s important to watch out for this whenever your code deals with truly large numbers, especially when raising numbers to arbitrary powers. I’ve seen this occur in backoff methods like the following:
def backoff_time(retry_count, initial_backoff, max_backoff): return min(max_backoff, initial_backoff*(2**retry_count))
It will work fine as along as
initial_backoff is integral, but if someone wants to have an
initial_backoff < 1, you’re at risk of overflow.
Want to get emailed about new blog posts?
Do you want to hire me? I’m available for engagements from 1 week to a few months. Hire me!