Pickle and savez_compressed compressed volume comparison

Sometimes, we need to transmit some numpy.ndarray objects in the network. At this time, we need to find an efficient serialization method suitable for network transmission:

from mark import BASE_DIR
from numpy import ndarray
import numpy
import pickle

img1 = numpy.array([i for i in range(5120)])


# 方式一： 使用 pickle 序列化 numpy.ndarray
with open(BASE_DIR/'testing'/'001.bin', 'wb') as f:
    img1: ndarray
    f.write(pickle.dumps(img1))

# --- 结果 41111 bytes

# 方法二： 使用 pickle 序列化 python 的 list 对象
with open(BASE_DIR/'testing'/'002.bin', 'wb') as f:
    ins: list[int] = img1.tolist()
    f.write(pickle.dumps(ins))

# --- 结果 15130 bytes

# 方法三： 使用 numpy 的 savez 序列化 numpy.ndarray
numpy.savez(BASE_DIR/'testing'/'numpy_savez_test', img1)

# --- 结果 41224 bytes

# 方法四： 使用 numpy 的 savez_compressed (带压缩功能) 序列化 numpy.ndarray
numpy.savez_compressed(BASE_DIR/'testing'/'numpy_savez_compressed_test', img1)

numpy.load(BASE_DIR/'testing'/'numpy_savez_compressed_test')

# --- 结果 7982 bytes


""" 

"""

in conclusion

结论：
方法四 < 方法二 < 方法一 < 方法三
(7982)< (15130)< (41111)< (41224)