Comment by feral | Hacker News Reader

Comment by feral | Hacker News Reader

Parent Full thread

I agree with the spirit of your argument.

In fact, I almost came here to write a parallel comment: I'm really not sure that reversing the list 'a' with 'a[::-1]' is better than 'reversed(a)', which usually effectively does the same thing, but whose meaning is much more obvious.

But, while I agree with your general point, in the specific case of 'defaultdict', I differ.

I use 'defaultdict' all the time and I'm glad its there. It feels cleaner than 'freqs.get(c,0)'. I define the default value in one place, and then the interface to my datastructure is simpler; hence as I continue writing, I can spend more of my brainpower in the problem domain.

Its a small detail, but its one less thing to think about when writing a complex algorithm.

I define the default value in one place, and then the interface to my datastructure is simpler; hence as I continue writing, I can spend more of my brainpower in the problem domain.

Actually, this speaks to point: optimization should be for reading, not for writing.

reversed(a) and a[::-1] are not equivalent. The former produces an iterator over the given list (with all the mutability dangers that come with it), while the latter produces a copied list. For plain iteration, you're correct, reversed() is better (similar to how xrange vs. range was back in the day); however, for reversing something and keeping it around, the slice syntax is better.

    >>> reversed([1, 2, 3])
    <listreverseiterator object at 0x107265650>
    >>> [1, 2, 3][::-1]
    [3, 2, 1]

To clarify your point, list(reversed(a)) and a[::-1] are equivalent. It's a slightly subtle point, but extremely important if you're keeping the result of reversed() around for any length of time. If you're just iterating at the moment that you use it, yes, they're effectively equivalent.

list(reversed(a))

Reply on news.ycombinator.com