I often find it cleaner to write a generator than to return a list. For example, I prefer

def my_func_gen(foo):
    for i in foo:
        # Do some stuff that's too complicated for a list or generator comprehension
        yield whatever

to

def my_func_list(foo):
    result = []
    for i in foo:
        # Do some stuff that's too complicated for a list or generator comprehension
        result.append(whatever)
    return result

Further, this answer says, "You're encouraged to use iterators for everything." So is my_func_gen better, then? Maybe not. I'd like for the caller not to even know whether it's getting a list or an iterator. But the caller has to think about it because, for example, the iterator won't be sliceable using Python's nice clean syntax; I'll have to use itertools.islice.

So what should I do?

1. Ignore the advice to "use iterators for everything"
2. Get out of the habit of using slices and other behavior that's not available to all iterables
3. Run list(my_iterable) any time I might want to use features that lists support and generators do not
4. Use lists sometimes and generators others. But when?