What is better? For-loops or generator expressions (in cases like this)

Question

Which version of this code is better?

Ver 1.

a = 0
for i in xrange( 1000 ):
    if i % 3 == 0 or i % 5 == 0:
        a += i
print a

Ver 2.

from operator import add
print reduce( add,
              ( i for i in xrange( 1000 )
                if i % 3 == 0 or i % 5 == 0 ) )

Things I like about version one is that it is more readable, but it creates requires a variable. If you had to choose, which one would you argue is better?

Answer 1

The second version can (and should) be written using sum instead of reduce(add,...), which is somewhat more readable:

sum(i for i in xrange( 1000 ) if i % 3 == 0 or i % 5 == 0)

That is the version that I'd prefer as far as implementations of this particular algorithm go because it's more declarative than a loop and just reads nicer in my opinion. (It's also faster though that's less of a concern).

However it should be noted that for summing all numbers up to n that divisible by 3 or 5, there's actually a better algorithm than going over all those numbers and summing them.

Answer 2

Don't use reduce, it is removed python 3.

kracekumar@python-lover:~$ python -m timeit 'sum(i for i in xrange(1000) if i%3 ==0 or i%5==0) '
1000 loops, best of 3: 232 usec per loop

kracekumar@python-lover:~$ python -m timeit 'sum([i for i in xrange(1000) if i%3 ==0 or i%5==0]) '

10000 loops, best of 3: 177 usec per loop

First method uses generators second one uses list comprehensions.

You can use both but generators will occupy less memory.

In [28]: g = (i for i in xrange(1000) if i%3 ==0 or i%5==0)

In [29]: l = [i for i in xrange(1000) if i%3 ==0 or i%5==0]

In [30]: import sys

In [31]: sys.getsizeof(l)
Out[31]: 2136

In [32]: sys.getsizeof(g)
Out[32]: 36

It is advised to use generators, here is a beautiful tutorial about generators.