pickle size efficiency for Pandas dataframes

picklesize.py

import numpy as np
import pickle

def pickle_size(obj):
    return len(pickle.dumps(obj))


def arr_size(arr):
    try:
        categories, codes = arr.categories, arr.codes
    except AttributeError:
        pass
    else:
        # Categorical.
        return arr_size(categories) + arr_size(codes)
    if arr.dtype == np.dtype(object):
        return sum( len(x) for x in arr )
    else:
        return arr.size * arr.itemsize


def data_size(df):
    cols = [ df[n] for n in df.columns ] + [df.index]
    return sum( arr_size(c.values) for c in cols )


def pickle_eff(df):
    return pickle_size(df) / data_size(df)