Is it safe to use next within a for loop in Python?

Question

Consider the following Python code:

b = [1,2,3,4,5,6,7]
a = iter(b)
for x in a :
    if (x % 2) == 0 :
        print(next(a))

Which will print 3, 5, and 7. Is the use of next on the variable being iterated on a reliable construct (you may assume that a StopIteration exception is not a concern or will be handled), or does the modification of the iterator being looped over inside the loop constitute a violation of some principle?

Answer 1

There's nothing wrong here protocol-wise or in theory that would stop you from writing such code. An exhausted iterator it will throw StopIteration on every subsequent call to it.__next__, so the for loop technically won't mind if you exhaust the iterator with a next/__next__ call in the loop body.

I advise against writing such code because the program will be very hard to reason about. If the scenario gets a little more complex than what you are showing here, at least I would have to go through some inputs with pen and paper and work out what's happening.

In fact, your code snippet possibly does not even behave like you think it behaves, assuming you want to print every number that is preceded by an even number.

>>> b = [1, 2, 4, 7, 8]                                              
>>> a = iter(b)                                                      
>>> for x in a: 
...:    if x%2 == 0: 
...:        print(next(a, 'stop'))                                   
4
stop

Why is 7 skipped although it's preceded by the even number 4?

>>>> a = iter(b)                                                      
>>>> for x in a: 
...:     print('for loop assigned x={}'.format(x)) 
...:     if x%2 == 0: 
...:         nxt = next(a, 'stop') 
...:         print('if popped nxt={} from iterator'.format(nxt)) 
...:         print(nxt)
...:                                               
for loop assigned x=1
for loop assigned x=2
if popped nxt=4 from iterator
4
for loop assigned x=7
for loop assigned x=8
if popped nxt=stop from iterator
stop

Turns out x = 4 is never assigned by the for loop because the explicit next call popped that element from the iterator before the for loop had a chance to look at the iterator again.

That's something I'd hate to work out the details of when reading code.

---

If you want to iterate over an iterable (including iterators) in "(element, next_element)" pairs, use the pairwise recipe from the itertools documentation.

from itertools import tee                                         

def pairwise(iterable):
    "s -> (s0,s1), (s1,s2), (s2, s3), ..." 
    a, b = tee(iterable) 
    next(b, None) 
    return zip(a, b) 

Demo:

>>> b = [1,2,3,4,5,6,7]                                               
>>> a = iter(b)                                                       
>>>                                                                   
>>> for x, nxt in pairwise(a): # pairwise(b) also works 
...:    print(x, nxt)                                                                      
1 2
2 3
3 4
4 5
5 6
6 7

In general, itertools together with its recipes provides many powerful abstractions for writing readable iteration-related code. Even more useful helpers can be found in the more_itertools module, including an implementation of pairwise.

Answer 2

It depends what you mean by 'safe', as others have commented, it is okay, but you can imagine some contrived situations that might catch you out, for example consider this code snippet:

b = [1,2,3,4,5,6,7]
a = iter(b)
def yield_stuff():
    for item in a:
        print(item)
        print(next(a))
    yield 1

list(yield_stuff())

On Python <= 3.6 it runs and outputs:

1
2
3
4
5
6
7

But on Python 3.7 it raises RuntimeError: generator raised StopIteration. Of course this is expected if you read PEP 479 and if you're thinking about handling StopIteration anyway you might never encounter it, but I guess the use cases for calling next() inside a for loop are rare and there are normally clearer ways of re-factoring the code.