Optimize unsafe_sample

[eendebakpt]

In unsafe_sample the sample_times is guaranteed to be a numpy array. Calling searchsorted from the array is more efficient.

For a small numpy array:

In [66]: %timeit sample_times.searchsorted( (0,0), 'left')
1.53 µs ± 63.9 ns per loop (mean ± std. dev. of 7 runs, 1,000,000 loops each)

In [67]: %timeit np.searchsorted(sample_times, (0,0), 'left')
2.73 µs ± 424 ns per loop (mean ± std. dev. of 7 runs, 100,000 loops each)

@terrorfisch

[github-actions]

Test Results

       6 files         6 suites   6m 50s ⏱️
1 216 tests 1 161 ✔️   55 💤 0 ❌
7 296 runs  6 966 ✔️ 330 💤 0 ❌

Results for commit 58a3471.

[terrorfisch]

Thanks. Seems like numpy's array_function_dispatch introduces quite some overhead.