pickle是Python标准库中的一个二进制序列化和反序列化库。
可以以二进制的形式将数据持久化保存到磁盘文件中。可以将数据和代码分离，提高代码可读性和优雅度。
一、pickle模块介绍
pickle模块可以对多种Python对象进行序列化和反序列化，序列化称为pickling，反序列化称为unpickling。
序列化是将Python对象转化为二进制数据，可以配合文件操作将序列化结果保存到文件中(也可以配合数据库操作保存到数据库中)。
反序列化则是将二进制数据还原回Python对象，先从文件中(或数据库中)读取出保存的二进制数据。
pickle模块常用的方法如下：
dump(obj, file): 将Python对象序列化，并将序列化结果写入到打开的文件中。
load(file): 从打开的文件中读取出保存的数据，将数据反序列化成Python对象。
dumps(obj): 将Python对象序列化，并直接返回序列化的二进制数据(类型为bytes)，而不写入文件。
loads(data): 将字节码数据反序列化成Python对象，传入的数据必须为二进制数据(bytes-like object)。
dump()和load()是互逆的方法，dumps()和loads()是互逆的方法，使用哪一对方法取决于是否要读写文件。
二、pickle可以序列化哪些Python对象
pickle与json相比，json数据有严格的格式要求，只能保存满足格式要求的数据，而pickle几乎可以序列化Python中的所有数据对象。
pickle可以序列化的Python对象如下：
None、True和False
整数、浮点数、复数
str、byte、bytearray
只包含可序列化对象的集合，包括tuple、list、set和dict
定义在模块最外层的函数(使用def定义，lambda函数不可以)
定义在模块最外层的内置函数
定义在模块最外层的类
某些类实例
三、案例分享
之前写过一篇使用matplotlib绘制柱状图的文章，参考：https://blog.csdn.net/weixin_43790276/article/details/109564348。
文章里有一个56行的字典，本文利用pickle模块来将字典序列化写入文件中，绘图时从文件中读取数据并反序列化，实现数据与代码的分离。
1. 将数据序列化保存
创建一个Python脚本pickling.py，先将56行的字典序列化保存。
# coding=utf-8
import pickle
data = {
"DWG-DRX1": [[(3, 2, 4), (2, 0, 4), (1, 0, 1), (3, 1, 4), (0, 0, 4)],
[(2, 3, 1), (0, 2, 1), (1, 0, 0), (0, 2, 1), (0, 2, 2)]],
"DWG-DRX2": [[(1, 2, 8), (6, 1, 5), (2, 1, 8), (3, 1, 7), (0, 2, 7)],
[(3, 3, 1), (0, 2, 5), (1, 3, 4), (2, 2, 4), (1, 2, 4)]],
"DWG-DRX3": [[(2, 2, 10), (7, 0, 6), (5, 0, 8), (3, 1, 6), (4, 4, 4)],
[(3, 4, 0), (2, 6, 2), (1, 3, 0), (1, 3, 3), (0, 5, 3)]],
"SN-JDG1": [[(4, 2, 9), (3, 1, 9), (5, 1, 11), (7, 3, 10), (1, 6, 7)],
[(3, 5, 8), (1, 5, 7), (2, 5, 7), (7, 2, 6), (0, 3, 10)]],
"SN-JDG2": [[(7, 2, 12), (7, 2, 14), (2, 0, 16), (9, 0, 12), (1, 4, 13)],
[(2, 6, 2), (2, 6, 4), (0, 4, 7), (4, 4, 1), (0, 6, 7)]],
"SN-JDG3": [[(5, 1, 5), (5, 1, 9), (3, 1, 8), (3, 1, 7), (1, 3, 11)],
[(0, 4, 2), (1, 2, 4), (0, 4, 3), (3, 1, 4), (3, 6, 3)]],
"SN-JDG4": [[(2, 2, 4), (3, 2, 5), (1, 0, 10), (7, 1, 5), (0, 2, 12)],
[(2, 3, 1), (2, 3, 3), (1, 3, 4), (0, 2, 6), (2, 2, 3)]],
"TES-FNC1": [[(2, 3, 8), (4, 2, 6), (2, 0, 8), (6, 0, 8), (1, 0, 10)],
[(0, 3, 3), (1, 3, 3), (4, 0, 0), (0, 6, 2), (0, 3, 3)]],
"TES-FNC2": [[(0, 2, 10), (8, 1, 4), (4, 0, 6), (4, 1, 5), (1, 2, 13)],
[(3, 2, 3), (1, 4, 5), (1, 2, 3), (0, 2, 6), (1, 7, 1)]],
"TES-FNC3": [[(3, 1, 4), (3, 1, 9), (3, 1, 7), (7, 1, 2), (0, 2, 12)],
[(0, 4, 3), (2, 6, 4), (2, 3, 2), (2, 0, 4), (0, 3, 3)]],
"TES-FNC4": [[(1, 2, 7), (10, 1, 7), (6, 2, 5), (0, 4, 16), (1, 4, 12)],
[(2, 3, 3), (3, 1, 5), (1, 4, 8), (4, 3, 5), (3, 7, 5)]],
"TES-FNC5": [[(1, 2, 1), (4, 1, 6), (4, 0, 6), (4, 1, 5), (0, 1, 6)],
[(2, 2, 1), (2, 3, 1), (0, 4, 1), (0, 1, 2), (0, 3, 2)]],
"G2-GEN1": [[(4, 0, 7), (2, 2, 11), (4, 1, 11), (6, 1, 6), (3, 0, 10)],
[(0, 5, 2), (3, 4, 1), (1, 3, 2), (0, 4, 1), (0, 3, 2)]],
"G2-GEN2": [[(3, 3, 14), (4, 3, 12), (11, 0, 11), (9, 2, 13), (1, 3, 15)],
[(3, 8, 1), (2, 5, 3), (2, 6, 5), (4, 4, 2), (0, 5, 7)]],
"G2-GEN3": [[(2, 5, 11), (7, 2, 10), (6, 3, 13), (7, 3, 11), (1, 1, 18)],
[(4, 5, 8), (2, 6, 7), (5, 4, 6), (3, 2, 6), (0, 6, 7)]],
"DWG-G21": [[(4, 0, 12), (7, 2, 9), (4, 2, 11), (6, 0, 9), (1, 2, 8)],
[(1, 5, 1), (3, 5, 2), (2, 5, 3), (0, 2, 3), (0, 5, 4)]],
"DWG-G22": [[(4, 2, 7), (5, 1, 9), (6, 2, 11), (7, 3, 9), (3, 1, 11)],
[(0, 7, 1), (0, 4, 4), (4, 4, 2), (3, 4, 1), (1, 6, 2)]],
"DWG-G23": [[(3, 1, 9), (6, 2, 5), (5, 2, 6), (8, 2, 7), (0, 3, 13)],
[(1, 3, 3), (3, 3, 4), (1, 4, 3), (2, 3, 3), (3, 9, 4)]],
"DWG-G24": [[(5, 0, 3), (2, 0, 7), (2, 0, 10), (2, 1, 3), (4, 1, 4)],
[(0, 5, 1), (1, 3, 0), (0, 3, 1), (1, 2, 1), (0, 2, 1)]],
"SN-TES1": [[(5, 1, 5), (3, 1, 6), (1, 0, 4), (2, 3, 3), (0, 2, 3)],
[(2, 4, 0), (0, 1, 4), (1, 2, 2), (4, 2, 0), (0, 2, 4)]],
"SN-TES2": [[(5, 1, 4), (1, 2, 5), (3, 1, 7), (3, 3, 4), (0, 0, 7)],
[(2, 1, 2), (1, 3, 5), (2, 5, 4), (2, 2, 0), (0, 1, 5)]],
"SN-TES3": [[(3, 0, 7), (2, 2, 4), (2, 1, 4), (5, 2, 4), (1, 2, 7)],
[(0, 3, 3), (2, 3, 3), (3, 1, 1), (0, 4, 4), (2, 2, 2)]],
"SN-TES4": [[(5, 2, 4), (1, 3, 16), (8, 1, 8), (6, 4, 9), (1, 8, 13)],
[(1, 2, 10), (9, 5, 4), (1, 4, 9), (5, 6, 10), (2, 4, 12)]],
"DWG-SN1": [[(2, 2, 11), (5, 3, 9), (8, 1, 11), (4, 2, 12), (2, 4, 7)],
[(1, 5, 5), (5, 4, 4), (3, 3, 2), (2, 3, 3), (1, 6, 3)]],
"DWG-SN2": [[(10, 1, 4), (2, 1, 10), (3, 3, 11), (3, 3, 10), (2, 4, 7)],
[(0, 4, 8), (5, 4, 2), (5, 6, 2), (2, 3, 5), (0, 3, 9)]],
"DWG-SN3": [[(3, 3, 10), (5, 2, 8), (3, 3, 3), (5, 1, 6), (0, 2, 8)],
[(3, 6, 5), (1, 2, 2), (4, 3, 2), (2, 3, 3), (1, 2, 6)]],
"DWG-SN4": [[(2, 0, 12), (8, 0, 7), (1, 3, 5), (9, 1, 5), (4, 3, 4)],
[(2, 9, 1), (1, 5, 2), (2, 2, 0), (2, 4, 2), (0, 4, 3)]],
}
with open('S10.pkl', 'wb') as pkl_file:
pickle.dump(data, pkl_file)序列化只需要两行代码，打开一个文件对象，使用dump()方法将字典序列化保存到了S10.pkl文件中，文件默认在代码的同级目录下，也可以指定绝对路径。注意，打开文件对象时使用wb模式。
S10.pkl的后缀名可以自定义(后缀名不会改变文件保存的格式)，不过因为是用pickle模块序列化的数据，通常都以.pkl作为后缀，方便识别。
直接用IDE打开文件S10.pkl，显示的内容是乱码的，因为保存的内容是二进制数据。
2. 读取数据并反序列化
# coding=utf-8
import matplotlib.pyplot as plt
from matplotlib import ticker
from numpy import mean
import pickle
with open('S10.pkl', 'rb') as pkl_file:
data = pickle.load(pkl_file)
location = ["上单", "打野", "中单", "下路", "辅助"]
win_loc_kill, win_loc_die, win_loc_assists = [[list() for _ in range(5)] for _ in range(3)]
lose_loc_kill, lose_loc_die, lose_loc_assists = [[list() for _ in range(5)] for _ in range(3)]
for i in range(5):
win_loc_kill[i] = [value[0][i][0] for value in data.values()]
win_loc_die[i] = [value[0][i][1] for value in data.values()]
win_loc_assists[i] = [value[0][i][2] for value in data.values()]
lose_loc_kill[i] = [value[1][i][0] for value in data.values()]
lose_loc_die[i] = [value[1][i][1] for value in data.values()]
lose_loc_assists[i] = [value[1][i][2] for value in data.values()]
# noinspection PyTypeChecker
win_avg_kill = [round(mean(kill), 2) for kill in win_loc_kill]
# noinspection PyTypeChecker
win_avg_die = [round(mean(die), 2) for die in win_loc_die]
# noinspection PyTypeChecker
win_avg_assists = [round(mean(assists), 2) for assists in win_loc_assists]
# noinspection PyTypeChecker
lose_avg_kill = [round(mean(kill), 2) for kill in lose_loc_kill]
# noinspection PyTypeChecker
lose_avg_die = [round(mean(die), 2) for die in lose_loc_die]
# noinspection PyTypeChecker
lose_avg_assists = [round(mean(assists), 2) for assists in lose_loc_assists]
fig, axs = plt.subplots(nrows=2, ncols=1, figsize=(20, 16), dpi=100)
x = range(len(location))
axs[0].bar([i-0.2 for i in x], win_avg_kill, width=0.2, color='b')
axs[0].bar(x, win_avg_die, width=0.2, color='r')
axs[0].bar([i+0.2 for i in x], win_avg_assists, width=0.2, color='g')
axs[1].bar([i-0.2 for i in x], lose_avg_kill, width=0.2, color='b')
axs[1].bar(x, lose_avg_die, width=0.2, color='r')
axs[1].bar([i+0.2 for i in x], lose_avg_assists, width=0.2, color='g')
for a, b in zip(x, win_avg_kill):
axs[0].text(a-0.2, b+0.1, '%.02f' % b, ha='center', va='bottom', fontsize=14)
for a, b in zip(x, win_avg_die):
axs[0].text(a, b+0.1, '%.02f' % b, ha='center', va='bottom', fontsize=14)
for a, b in zip(x, win_avg_assists):
axs[0].text(a+0.2, b+0.1, '%.02f' % b, ha='center', va='bottom', fontsize=14)
for a, b in zip(x, lose_avg_kill):
axs[1].text(a-0.2, b+0.1, '%.02f' % b, ha='center', va='bottom', fontsize=14)
for a, b in zip(x, lose_avg_die):
axs[1].text(a, b+0.1, '%.02f' % b, ha='center', va='bottom', fontsize=14)
for a, b in zip(x, lose_avg_assists):
axs[1].text(a+0.2, b+0.1, '%.02f' % b, ha='center', va='bottom', fontsize=14)
for i in range(2):
axs[i].xaxis.set_major_locator(ticker.FixedLocator(x))
axs[i].xaxis.set_major_formatter(ticker.FixedFormatter(location))
axs[i].set_yticks(range(0, 11, 2))
axs[i].grid(linestyle="--", alpha=0.5)
axs[i].legend(['击杀', '死亡', '助攻'], loc='upper left', fontsize=16, markerscale=0.5)
axs[i].set_xlabel("位置", fontsize=18)
axs[i].set_ylabel("场均数据", fontsize=18, rotation=0)
axs[0].set_title("S10总决赛胜方各位置场均数据", fontsize=18)
axs[1].set_title("S10总决赛负方各位置场均数据", fontsize=18)
plt.show()
反序列化代码也只有两行，打开文件S10.pkl，然后使用load()方法对文件对象反序列化，返回数据。打开文件对象时使用rb模式。
运行代码，绘图功能正常。
经过pickle模块的序列化和反序列化，将数据持久化到了文件S10.pkl中。实现了数据与代码的分离，避免了直接在代码中写一个很长的字典数据，代码更加优雅。
在上面的例子中，对一个56行的数据进行序列化，已经有不错的效果了。在实际的项目中，数据更大，将数据放到代码中会占很大的篇幅，进行序列化处理的优化效果会更明显。
如果有多个脚本使用同一份数据，可以直接读取同一个序列化数据文件，避免了在不同脚本中粘贴同一份数据。
以上就是pickle模块的序列化介绍，如果需要本文代码，可以扫码关注公众号“Python碎片”，然后在后台回复“pickle”关键字获取完整代码。
