feather_speedtest.py

import timeit
import numpy as np
import pandas as pd
from pathlib import Path

def create_test_data(large=False, small=False):
    # Note that a large_number by 1 dataframe is not really representative.
    root = Path("tmp")
    root.mkdir(exist_ok=True)
    if large:
        large_df = pd.DataFrame(np.random.randint(0,100,size=(800000000,1)))
        large_df.columns = large_df.columns.astype(str)
        large_df.to_feather(root/"large_df.feather")
    if small:
        for i in range(20000):
            small_df = pd.DataFrame(np.random.randint(0,100,size=(40000,1)))
            small_df.columns = small_df.columns.astype(str)
            small_df.to_feather(root/f"small_df_{i}.feather")

def test_speed(large=False, small=False):
    if large:
        print(timeit.timeit("""pd.read_feather("tmp/large_df.feather")""", setup="import pandas as pd", number=10))
    if small:
        print(timeit.timeit("""for f in Path("tmp").glob("small_*.feather"): pd.read_feather(f)""", setup="import pandas as pd; from pathlib import Path", number=10))

if __name__ == '__main__':
    create_test_data(large=False, small=True)
    test_speed(large=False, small=True)
    # Loading a single 6 GB file takes roughly 15.2 seconds
    # In contrast to loading 20,000 300kb files, which takes roughly forever (296.8 seconds).
    # (Ten times average)

Nope, I ran it as is i.e number=10. only changed the boolean values