python 深拷贝和浅拷贝（shallow and deep copy）

1，深拷贝和浅拷贝只相对于容器类型的对象（compound obejects）来说，对于原子类型的对象（atomic s）没有这个概念。看下面的代码：

num = 123
num_c = num
print(id(num), id(num_c))
num_c = 456
print(id(num), id(num_c))
# 输出为
# 4526750784 4526750784
# 4526750784 4563542224

string = \'abc\'
string_c = string
print(id(string), id(string_c))
string_c = \'def\'
print(id(string), id(string_c))
# 输出为
# 4528361176 4528361176
# 4528361176 4529762464

对于原子类型，赋值（assignment）只是创建了一个引用指向另一个对象，改变其中一个对象不会影响另一个对象。

2，对于容器对象，正常赋值（normal assignment）和切片赋值（slice assignment）效果不一样。代码如下：

lis = [\'a\', \'b\', \'c\']
lis_c = lis
print(id(lis), id(lis_c))
[print(x, id(x)) for x in lis]
[print(x, id(x)) for x in lis_c]
print(\'After\')
lis_c[1:3] = [\'d\', \'e\']
print(id(lis), id(lis_c))
[print(x, id(x)) for x in lis]
[print(x, id(x)) for x in lis_c]
# 输出为
4564651592 4564651592
a 4528834072
b 4528276568
c 4528219168
a 4528834072
b 4528276568
c 4528219168
After
4564651592 4564651592
a 4528834072
d 4528456400
e 4528388952
a 4528834072
d 4528456400
e 4528388952

lis = [\'a\', \'b\', \'c\']
lis_c = lis[:]
print(id(lis), id(lis_c))
[print(x, id(x)) for x in lis]
[print(x, id(x)) for x in lis_c]
print(\'After\')
lis_c[1:3] = [\'d\', \'e\']
print(id(lis), id(lis_c))
[print(x, id(x)) for x in lis]
[print(x, id(x)) for x in lis_c]
# 输出为
4564225608 4564496008
a 4528834072
b 4528276568
c 4528219168
a 4528834072
b 4528276568
c 4528219168
After
4564225608 4564496008
a 4528834072
b 4528276568
c 4528219168
a 4528834072
d 4528456400
e 4528388952

lis = [\'a\', \'b\', \'c\']
lis_c = lis
print(id(lis), id(lis_c))
[print(x, id(x)) for x in lis]
[print(x, id(x)) for x in lis_c]
print(\'After\')
lis_c = [\'d\', \'e\', \'f\']
print(id(lis), id(lis_c))
[print(x, id(x)) for x in lis]
[print(x, id(x)) for x in lis_c]
# 输出为
4562685000 4562685000
a 4528834072
b 4528276568
c 4528219168
a 4528834072
b 4528276568
c 4528219168
After
4562685000 4564494984
a 4528834072
b 4528276568
c 4528219168
d 4528456400
e 4528388952
f 4527690168

可以看出，正常赋值lis_c和lis指向同样的内存地址。对于正常赋值，改变lis_c的部分元素，lis的元素也会跟着改变；而对于切片赋值，二者指向不同的地址，所以不会相互影响。对正常赋值，如果改变整个lis_c中的元素，则lis_c的内存地址会发生改变，所以lis_c和lis不会相互影响。

3，字典的性质与列表相似，具有2中的性质，改变字典部分的key或全部的key，字典的内存地址会变；改变部分的value或全部的value，字典的内存地址不变。

dic = {\'a\': 1, \'b\': 2, \'c\': 3}
dic_c = dic
print(id(dic), id(dic_c))
[print(\'key:\', x, id(x), \'value:\', y, id(y)) for x, y in dic.items()]
[print(\'key:\', x, id(x), \'value:\', y, id(y)) for x, y in dic_c.items()]
print(\'change part of key\')
dic_c = {\'aa\': 1, \'bb\': 2, \'c\': 3}
print(id(dic), id(dic_c))
[print(\'key:\', x, id(x), \'value:\', y, id(y)) for x, y in dic.items()]
[print(\'key:\', x, id(x), \'value:\', y, id(y)) for x, y in dic_c.items()]
# 输出为
4564146576 4564146576
key: a 4528834072 value: 1 4526746880
key: b 4528276568 value: 2 4526746912
key: c 4528219168 value: 3 4526746944
key: a 4528834072 value: 1 4526746880
key: b 4528276568 value: 2 4526746912
key: c 4528219168 value: 3 4526746944
change part of key
4564146576 4564735176
key: a 4528834072 value: 1 4526746880
key: b 4528276568 value: 2 4526746912
key: c 4528219168 value: 3 4526746944
key: aa 4550712208 value: 1 4526746880
key: bb 4543935296 value: 2 4526746912
key: c 4528219168 value: 3 4526746944

dic = {\'a\': 1, \'b\': 2, \'c\': 3}
dic_c = dic
print(id(dic), id(dic_c))
[print(\'key:\', x, id(x), \'value:\', y, id(y)) for x, y in dic.items()]
[print(\'key:\', x, id(x), \'value:\', y, id(y)) for x, y in dic_c.items()]
print(\'change all values\')
dic_c[\'a\'] = 11
dic_c[\'b\'] = 22
dic_c[\'c\'] = 33
print(id(dic), id(dic_c))
[print(\'key:\', x, id(x), \'value:\', y, id(y)) for x, y in dic.items()]
[print(\'key:\', x, id(x), \'value:\', y, id(y)) for x, y in dic_c.items()]
# 输出为
4564735320 4564735320
key: a 4528834072 value: 1 4526746880
key: b 4528276568 value: 2 4526746912
key: c 4528219168 value: 3 4526746944
key: a 4528834072 value: 1 4526746880
key: b 4528276568 value: 2 4526746912
key: c 4528219168 value: 3 4526746944
change all values
4564735320 4564735320
key: a 4528834072 value: 11 4526747200
key: b 4528276568 value: 22 4526747552
key: c 4528219168 value: 33 4526747904
key: a 4528834072 value: 11 4526747200
key: b 4528276568 value: 22 4526747552
key: c 4528219168 value: 33 4526747904

4，对于元祖来说，改变其中一个元素，就使整个元祖的内存地址发生改变。

tup = (\'a\', \'b\', [\'c\', \'d\'])
tup_c = tup
print(id(tup), id(tup_c))
[print(x, id(x)) for x in tup]
[print(x, id(x)) for x in tup_c]
print(\'After\')
tup_c = (\'a\', \'b\', [\'c\', \'e\'])
print(id(tup), id(tup_c))
[print(x, id(x)) for x in tup]
[print(x, id(x)) for x in tup_c]
# 输出为
4564433872 4564433872
a 4528834072
b 4528276568
[\'c\', \'d\'] 4564658312
a 4528834072
b 4528276568
[\'c\', \'d\'] 4564658312
After
4564433872 4564147008
a 4528834072
b 4528276568
[\'c\', \'d\'] 4564658312
a 4528834072
b 4528276568
[\'c\', \'e\'] 4564603272

5，通过下面三种方法可以建立浅拷贝： (1) taking a complete slice [:], (2) using a factory function, e.g., list(), dict(), etc., or (3) using the copy() function of the copy module。深拷贝要用到copy module中的copy.deepcopy()。

import copy
a = [\'123\', \'456\', [\'a\', \'b\']]
b = copy.copy(a)
print(id(a), id(b))
b[2][0] = [\'d\']
print(a, id(a), b, id(b))
# 输出为
4564727816 4564602952
[\'123\', \'456\', [[\'d\'], \'b\']] 4564727816 [\'123\', \'456\', [[\'d\'], \'b\']] 4564602952

import copy
a = [\'123\', \'456\', [\'a\', \'b\']]
b = copy.deepcopy(a)
print(id(a), id(b))
b[2][0] = [\'d\']
print(a, id(a), b, id(b))
# 输出为
4564493896 4564658824
[\'123\', \'456\', [\'a\', \'b\']] 4564493896 [\'123\', \'456\', [[\'d\'], \'b\']] 4564658824

可以看出，无论浅拷贝还是深拷贝，list在赋值前后的内存地址不一样，不同点是深拷贝会递归地为容器中的容器进行拷贝；浅拷贝只对最外层拷贝，内层的容器还是指向同一个地址，会相互影响。