Avoid islice overhead in itertoolz.second

[groutr]

Using itertools.islice seems to be slower than simply calling next twice. In each case, calling next twice is faster.

m0 = range(5)
m1 = range(100)
m2 = range(10000)

In [17]: %timeit second(m0)
298 ns ± 5.18 ns per loop (mean ± std. dev. of 7 runs, 1000000 loops each)

In [18]: %timeit second(m1)
296 ns ± 4.49 ns per loop (mean ± std. dev. of 7 runs, 1000000 loops each)

In [19]: %timeit second(m2)
290 ns ± 3.98 ns per loop (mean ± std. dev. of 7 runs, 1000000 loops each)
===============================================
In [21]: %timeit my_second(m0)
263 ns ± 6.78 ns per loop (mean ± std. dev. of 7 runs, 1000000 loops each)

In [22]: %timeit my_second(m1)
257 ns ± 2.12 ns per loop (mean ± std. dev. of 7 runs, 1000000 loops each)

In [23]: %timeit my_second(m3)
263 ns ± 6.2 ns per loop (mean ± std. dev. of 7 runs, 1000000 loops each)

[eriknw]

Nice. This is what cytoolz does as well. Should we add a try/except block as was needed for #409?

[groutr]

@eriknw, you're right, we need a try/except around the next() call. And interpose needs to be changed as well. I'll do that in another PR.

The try/except might negate the performance gain here though.

[groutr]

In [21]: %timeit second(m0) # Released implementation
385 ns ± 9.37 ns per loop (mean ± std. dev. of 7 runs, 1000000 loops each)
In [22]: %timeit second(m1)
382 ns ± 12 ns per loop (mean ± std. dev. of 7 runs, 1000000 loops each)
In [23]: %timeit second(m2)
378 ns ± 2.93 ns per loop (mean ± std. dev. of 7 runs, 1000000 loops each)
In [25]: %timeit my_second(m0)  # Final version with try/except
365 ns ± 7.99 ns per loop (mean ± std. dev. of 7 runs, 1000000 loops each)
In [26]: %timeit my_second(m1)
359 ns ± 4.69 ns per loop (mean ± std. dev. of 7 runs, 1000000 loops each)
In [27]: %timeit my_second(m2)
362 ns ± 5.31 ns per loop (mean ± std. dev. of 7 runs, 1000000 loops each)

[groutr]

@eriknw, testing on Python 3.7, I don't think second needs the fix from #409 because second is not a generator. I'm reverting the last commit.

[eriknw]

heh, my bad. LGTM. Thanks @groutr!

[groutr]

@eriknw no worries. It made me read the PEP a little more thoroughly.