Writing a Pythonic prime sieve

Question

I'm practicing writing code less by brute force. I'd like feedback for writing a prime sieve more Pythonic, simpler, or more creatively perhaps.

def prime_sieve(limit):
 numbers = dict()
 limit = int(limit)
 for n in range(2, limit + 1):
 numbers[n] = ()
 primes = numbers.copy()
 for n in numbers:
 m = n
 while m <= limit:
 m = m + n
 primes.pop(m, 0)
 return primes

Answer 1

Your implementation of the sieve, using a dictionary, throws away many of the benefits of the approach; dictionaries are unordered. Even a naive list-based implementation ends up processing fewer than half as many numbers (146 vs. 382 with a limit of 100):

def sieve(limit):
 nums = range(limit+1)
 nums[1] = 0
 for n in nums:
 if n:
 for x in range(2*n, limit+1, n):
 nums[x] = 0 # counting this vs. pop
 return set(filter(None, primes))

Your approach processes multiples of numbers that will subsequently turn out to be non-prime themselves, e.g. you might pop all multiples of 9, then go back over them when you later process 3.

Note that I have used a set, which is effectively a dictionary with no values, rather than have some dummy value.

Also, there was no need to copy the whole dictionary; for n in numbers.keys() would have worked fine.

There are numerous other implementations, in Python and other languages, on this site and elsewhere; I suggest you look at a few to see what other alterations could be made.