Longest consecutive region meeting a specified predicate (in `pandas`)

region.py

#!/usr/bin/env python3


from numpy.random import default_rng
from pandas import Series, date_range, MultiIndex, Index, to_timedelta
from string import ascii_lowercase

rng = default_rng(0)

s = Series(
    index=(idx := MultiIndex.from_product([
        rng.choice([*ascii_lowercase], size=(100, 4)).view('<U4').ravel(),
        date_range('2020-01', freq='T', periods=180*24*60),
    ], names=['entity', 'date'])),
    data=rng.integers(-10, +10, size=len(idx)),
    name='signal',
).pipe(
    lambda s: s.set_axis(
        MultiIndex.from_arrays([
            s.index.get_level_values('entity'),
            s.index.get_level_values('date') + to_timedelta(rng.integers(3600, size=len(s)), unit='s'),
        ])
    )
    .sample(frac=.50)
    .sort_index()
)

def consecutive_regions(obj, pred):
    return Series(
        data=(pred != pred.shift()).cumsum(),
        index=obj.index,
        name='region',
    )

print(
    s.groupby('entity').apply(
        lambda g:
            (reg := consecutive_regions(g, pred := g > 0)).loc[pred]
                .pipe(lambda s: s.loc[s == s.groupby(s).count().idxmax()])
                .pipe(lambda s: g.loc[s.index].droplevel('entity'))
    )
    sep='\n',
)