Speed up audinterface.Segment

[hagenw]

Similar to #100 we want to speed up audinterface.Segment when processing several files.

At the moment we get:

import pandas as pd

import audb
import audinterface
import audiofile


db = audb.load(
    'emodb',
    version='1.3.0',
    format='wav',
    sampling_rate=16000,
    mixdown=True,
)
files = db.files

def process_func(x, sr):
    return pd.MultiIndex.from_tuples(
        [(pd.to_timedelta(0), pd.to_timedelta(1))],
        names=['start', 'end'],
    )

# slow

segment = audinterface.Segment(
    process_func=process_func,
)

t = time.time()
index_slow = segment.process_files(files)
print(time.time() - t)

# fast

t = time.time()
starts = []
ends = []
for idx, file in enumerate(files):
    signal, sampling_rate = audiofile.read(file)
    index = process_func(
        signal,
        sampling_rate,
    )
    starts += list(index.get_level_values('start'))
    ends += list(index.get_level_values('end'))

index_fast = pd.MultiIndex.from_arrays(
    [files, starts, ends],
    names=['file', 'start', 'end'],
)
print(time.time() - t)

pd.testing.assert_index_equal(index_slow, index_fast)

4.128075122833252                                                                                   
1.7938168048858643