News for hackers. Stuff that matters.
Re:Python’s tug of war between beginner-friendly and advance... by squaresmile on Thursday February 04, 2021 @08:21PM (#26031937)
Interesting, I'm using this [0] to test and got different results. The Pythonic function is 2% slower on 3.9.1 AMD 4900HS but 17% faster on 3.9.1 E5-2660 v2 VPS and 6% faster on 3.6.9 i5-3570S bare metal.

I personally think either function is fine. The max finding code pattern in the long function should be familiar to most. For the Pythonic function, I don't think it's less obvious enough to matter. You have a list of stuffs and you want to find the max so you wrap it in the max function. Maybe zip(xs, xs[1:]) can take a bit to recognize but I think it's eh. For the Pythonic one, one can also write

    all_diffs = [x2-x1 for x1, x2 in zip(xs, xs[1:])]
    return max(all_diffs)
Using list is also markedly faster than the Pythonic generator for me, which is to be expected: ~16% faster than both.

[0] https://gist.github.com/squaresmile/8dd08898851a4e19359cdb30...

Re:Python’s tug of war between beginner-friendly and advance... by ALittleLight on Thursday February 04, 2021 @08:59PM (#26032189)
That is interesting. Trying your timeit code I've come to see that maybe something was wrong about how I was estimating the different performance. I was just using cProfile and profiling the two functions with big inputs and then comparing the cumulative time results that way. When I use your timeit examples my results (Intel Core i7-6700K CPU @ 4.00GHz) are different and closer to yours (pythonic ~1% slower). Thanks for sharing.