The Cost of Dynamism (was Re: Pyhon 2.x or 3.x, which is faster?)
Chris Angelico
rosuav at gmail.com
Sun Mar 13 16:57:31 EDT 2016
On Mon, Mar 14, 2016 at 6:39 AM, BartC <bc at freeuk.com> wrote:
> I used it in my benchmark to replace the if-else chain checking three lots
> of ranges:
>
> switch(c)
> if case(ord("A"),ord("B"),ord("C"),ord("D"),ord("E"),ord("F"),
> ord("G"),ord("H"),ord("I"),ord("J"),ord("K"),ord("L"),
> ord("M"),ord("N"),ord("O"),ord("P"),ord("Q"),ord("R"),
> ord("S"),ord("T"),ord("U"),ord("V"),ord("W"),ord("X"),
> ord("Y"),ord("Z")):
> upper+=1
> elif case(ord("a"),ord("b"),ord("c"),ord("d"),ord("e"),ord("f"),
> ord("g"),ord("h"),ord("i"),ord("j"),ord("k"),ord("l"),
> ord("m"),ord("n"),ord("o"),ord("p"),ord("q"),ord("r"),
> ord("s"),ord("t"),ord("u"),ord("v"),ord("w"),ord("x"),
> ord("y"),ord("z")):
> lower+=1
> elif case(ord("0"),ord("1"),ord("2"),ord("3"),ord("4"),ord("5"),
> ord("6"),ord("7"),ord("8"),ord("9")):
> digits+=1
> else:
> other+=1
>
> It worked, but took 110 seconds; 80 seconds without the ord's and comparing
> strings (but I still think it's perverse that integer ops are slower than
> string ops).
>
> But 110 or 80 seconds, the original Python was 3.6 seconds. (Probably,
> someone could tweak it to work with ranges, but this is extra programmer
> effort that you say is too valuable to waste on such matters.)
This is not comparing ranges, though. This is comparing against
individual values. To talk about comparing ranges, I would expect the
code to look something like this:
switch(c)
if case("A", "Z"):
upper += 1
elif case("a", "z"):
lower += 1
elif case("0", "9"):
digits += 1
else:
other += 1
THIS is comparing ranges. The underlying comparisons must be
inequalities, not equalities. I absolutely *do not care* about
performance until the code looks good - at least reasonably good.
(By the way, your switch/case pair is non-reentrant. Plus it uses a
class in a weird way. Why not, if you're working like this, just have
two functions and a module-global? Just as non-reentrant, much cleaner
code.)
ChrisA
More information about the Python-list
mailing list