PERF: lib.generate_slices

[mzeitlin11]

import numpy as np
import pandas._libs.lib as lib

np.random.seed(0)
inds = np.random.randint(0, 100, 1000000)
ngroups = inds.max() + 1

%timeit lib.generate_slices(inds, ngroups)

Master:
1.86 ms ± 44 µs per loop (mean ± std. dev. of 7 runs, 1000 loops each)

This pr:
859 µs ± 127 µs per loop (mean ± std. dev. of 7 runs, 1000 loops each)

Adding the cython decorators makes a huge difference here

[jreback]

@meeseeksdev backport 1.3.x

[jreback]

thanks @mzeitlin11

[lumberbot-app]

Something went wrong ... Please have a look at my logs.

[jbrockmendel]

FYI there is a tiny perf bump from using starts.base instead of np.asarray(starts) in these cases.

[mzeitlin11]

Ah thanks, good to know