Consider using object instead of string in index

[hagenw]

It turns out that even in version 2.2.0 of pandas the new string dtype is not up to the same speed for some tasks, and unfortunately one of them is indexing:

import pandas as pd
import timeit

points = 1000000
data = [f"data-{n}" for n in range(points)]
for dtype in ["object", "string", "string[pyarrow]"]:
    index = pd.Index([f"index-{n}" for n in range(points)], dtype=dtype)
    df = pd.DataFrame(data, index=index, dtype=dtype)
    print(dtype)
    %timeit df.loc['index-2000']

which returns

object
9.78 µs ± 18.9 ns per loop (mean ± std. dev. of 7 runs, 100,000 loops each)
string
15.7 µs ± 36.5 ns per loop (mean ± std. dev. of 7 runs, 100,000 loops each)
string[pyarrow]
17.6 µs ± 66.6 ns per loop (mean ± std. dev. of 7 runs, 100,000 loops each)

So we might consider switching to store the file index again as object dtype as we do now for the dependencies in audb (audeering/audb#371). The only problem is, that in audb the change is hidden for the user, whereas here it would be a breaking change.