1. import os
    2. import time
    4. import numpy as np
    5. import pandas as pd
    7. data = np.random.standard_normal((10 ** 7, 4))
    8. df = pd.DataFrame(data)
    9. methods = [
    10.     'default',
    11.     "zlib",
    12.     "lzo",
    13.     "bzip2",
    14.     "blosc",
    15.     'blosc:blosclz',
    16.     "blosc:lz4",
    17.     "blosc:lz4hc",
    18.     "blosc:snappy",
    19.     "blosc:zlib",
    20.     "blosc:zstd",
    21. ]
    22. for method in methods:
    23.     level, comp = 9, method
    24.     if method == 'default':
    25.         level, comp = None, None
    26.     start_time = time.time()
    27.     file = f"./data/{method.replace(':', '-')}.h5"
    28.     df.to_hdf(file, 'df', complevel=level, complib=comp)
    29.     print(method, os.path.getsize(file), time.time() - start_time, sep='\t')

    性能比较：

    1. method                    size                  time
    2. -------------------------------------------------
    3. default            400007240          2.722801923751831
    4. zlib                283500043        166.24143624305725
    5. lzo                305977502          0.4248645305633545
    6. bzip2                291635531         35.23882055282593
    7. blosc                309459970           1.6485953330993652
    8. blosc:blosclz        309459994          1.662555456161499
    9. blosc:lz4            305701404          0.4089069366455078
    10. blosc:lz4hc        293242558         16.87893295288086
    11. blosc:snappy        324139336          0.3809819221496582
    12. blosc:zlib        281797972        164.94474864006042
    13. blosc:zstd        287126493         42.24786901473999