【python】 深拷貝和淺拷貝:copy模塊、可變對(duì)象和不可變對(duì)象、其他拷貝方式

Python賦值操作或函數(shù)參數(shù)傳遞傳遞的永遠(yuǎn)是對(duì)象引用(即內(nèi)存地址),而不是對(duì)象內(nèi)容。在Python中一切皆對(duì)象,對(duì)象又分為可變(mutable)不可變(immutable)兩種類型。

對(duì)象拷貝是指在內(nèi)存中創(chuàng)建新的對(duì)象,產(chǎn)生新的內(nèi)存地址。當(dāng)頂層對(duì)象和它的子元素對(duì)象全都是immutable不可變對(duì)象時(shí),不存在被拷貝,因?yàn)闆]有產(chǎn)生新對(duì)象。

  • 淺拷貝(Shallow Copy):拷貝頂層對(duì)象,但不會(huì)拷貝內(nèi)部的子元素對(duì)象。
  • 深拷貝(Deep Copy):遞歸拷貝頂層對(duì)象,以及它內(nèi)部的子元素對(duì)象

可變對(duì)象和不可變對(duì)象

對(duì)象的類型決定了它裝著的數(shù)據(jù)是允許被修改的變量(可變的mutable)還是不可被修改的常量(不可變的immutable)。你可以把不可變對(duì)象想象成一個(gè)透明但封閉的盒子:你可以看到里面裝的數(shù)據(jù),但是無法改變它。類似地,可變對(duì)象就像一個(gè)開著口的盒子,你不僅可以看到里面的數(shù)據(jù),還可以拿出來修改它,但你無法改變這個(gè)盒子本身,即你無法改變對(duì)象的類型。

mutable: 可變對(duì)象,如List、Dict

immutable: 不可變對(duì)象,如Number、String、Tuple、Frozenset

【注釋】:Python賦值操作或函數(shù)參數(shù)傳遞,傳遞的永遠(yuǎn)是對(duì)象引用(即內(nèi)存地址),而不是對(duì)象內(nèi)容


關(guān)于 copy模塊

對(duì)象拷貝:
是指在內(nèi)存中創(chuàng)建新的對(duì)象,產(chǎn)生新的內(nèi)存地址。

【特點(diǎn)如下】:
(1)淺拷貝只拷貝最外層對(duì)象,深拷貝還會(huì)遞歸拷貝內(nèi)層對(duì)象

(2)無論是淺拷貝還是深拷貝,只拷貝mutable可變對(duì)象成為一個(gè)新對(duì)象,而immutable不可變對(duì)象還是原來的那個(gè);

(3)當(dāng)頂層對(duì)象和它的子元素對(duì)象全都是immutable不可變對(duì)象時(shí),因?yàn)闆]有產(chǎn)生新對(duì)象,所以不存在被拷貝;


關(guān)于淺拷貝

【一句話介紹】淺拷貝(Shallow Copy),拷貝頂層對(duì)象,但不會(huì)拷貝內(nèi)部的子元素對(duì)象。

【換句話說】傳遞的是地址,不會(huì)新建一個(gè)對(duì)象。

分別討論以下情況:

  • (1)當(dāng)頂層對(duì)象是mutable可變對(duì)象,但是它的子元素對(duì)象全都是immutable不可變對(duì)象

程序環(huán)境基于【jupyter-notebook】

In [1]: a = [1, 'world', 2]

In [2]: [ id(item) for item in a ]
Out[2]: [9164864, 140104749066928, 9164896]

In [3]: id(a)
Out[3]: 140104759916040

【小提升】:導(dǎo)入copy模塊,使用copy.copy()函數(shù)淺拷貝a,并賦值給變量b。

In [4]: import copy

In [5]: b = copy.copy(a)

In [6]: b
Out[6]: [1, 'world', 2]

In [7]: [ id(item) for item in b ]
Out[7]: [9164864, 140104749066928, 9164896]

In [8]: id(b)
Out[8]: 140104760027784

  • (2)當(dāng)頂層對(duì)象是mutable可變對(duì)象,但子元素也存在mutable可變對(duì)象(子元素部分immutable)

【舉個(gè)栗子】:

In [1]: a = [1, 2, ['hello','world']]

In [2]: import copy

In [3]: b = copy.copy(a)

In [4]: id(a)
Out[4]: 139770596269064

In [5]: id(b)
Out[5]: 139770596639368

In [6]: [ id(item) for item in a ]
Out[6]: [9164864, 9164896, 139770596304840]

In [7]: [ id(item) for item in b ]
Out[7]: [9164864, 9164896, 139770596304840]

In [8]: [ id(item) for item in a[2] ]
Out[8]: [139770585378520, 139770585378408]

In [9]: [ id(item) for item in b[2] ]
Out[9]: [139770585378520, 139770585378408]

【解釋】:淺拷貝copy.copy()只拷貝了頂層對(duì)象,沒有拷貝子元素對(duì)象['hello','world'],即a[2]和b[2]指向同一個(gè)列表對(duì)象


  • (3)當(dāng)頂層對(duì)象是immutable不可變對(duì)象,同時(shí)它的子元素對(duì)象也全都是immutable不可變對(duì)象
In [1]: a = (1, 2, 3)

In [2]: import copy

In [3]: b =  copy.copy(a)

In [4]: id(a)
Out[4]: 139664680010016

In [5]: id(b)
Out[5]: 139664680010016

In [6]: [ id(item) for item in a ]
Out[6]: [9164864, 9164896, 9164928]

In [7]: [ id(item) for item in b ]
Out[7]: [9164864, 9164896, 9164928]

【解釋】:變量a與變量b指向的是同一個(gè)元組對(duì)象,沒有拷貝


  • (4)當(dāng)頂層對(duì)象是immutable不可變對(duì)象時(shí),但子元素存在mutable可變對(duì)象(子元素部分mutable)
    來個(gè)栗子;
In [1]: a = (1, 2, ['hello','world'])

In [2]: import copy

In [3]: b = copy.copy(a)

In [4]: id(a)
Out[4]: 139650704096640

In [5]: id(b)
Out[5]: 139650704096640

In [6]: [ id(item) for item in a ]
Out[6]: [9164864, 9164896, 139650704068680]

In [7]: [ id(item) for item in b ]
Out[7]: [9164864, 9164896, 139650704068680]

In [8]: [ id(item) for item in a[2] ]
Out[8]: [139650692293328, 139650692293216]

In [9]: [ id(item) for item in b[2] ]
Out[9]: [139650692293328, 139650692293216]

In [10]: a[2][1] = 'china'

In [11]: a
Out[11]: (1, 2, ['hello', 'china'])

In [12]: b
Out[12]: (1, 2, ['hello', 'china'])

【解釋一下】:變量a與變量b指向的是相同的元組對(duì)象,并且a[2]與b[2]指向同一個(gè)列表,所以修改a[2][1]會(huì)影響b[2][1]


關(guān)于深拷貝

【一句話介紹】:
深拷貝(Deep Copy),遞歸拷貝頂層對(duì)象,以及它內(nèi)部的子元素對(duì)象。
【換句話說】:深拷貝時(shí),會(huì)拷貝可變?cè)乇旧?/strong>,而不是地址。

  • (1)當(dāng)頂層對(duì)象是mutable可變對(duì)象,但是它的子元素對(duì)象全都是immutable不可變對(duì)象
In [1]: a = [1, 'world', 2]

In [2]: import copy

In [3]: b = copy.deepcopy(a)

In [4]: id(a)
Out[4]: 140664823442376

In [5]: id(b)
Out[5]: 140664823349192

In [6]: [ id(item) for item in a ]
Out[6]: [9164864, 140664823391544, 9164896]

In [7]: [ id(item) for item in b ]
Out[7]: [9164864, 140664823391544, 9164896]

In [8]: a[0] = 3

In [9]: a
Out[9]: [3, 'world', 2]

In [10]: b
Out[10]: [1, 'world', 2]

In [11]: [ id(item) for item in a ]
Out[11]: [9164928, 140664823391544, 9164896]

In [12]: [ id(item) for item in b ]
Out[12]: [9164864, 140664823391544, 9164896]

【解釋】:變量a與變量b指向不同的列表對(duì)象,修改a[0]只是將列表a的第一個(gè)元素重新指向新對(duì)象,不會(huì)影響b[0]


  • (2)當(dāng)頂層對(duì)象是mutable可變對(duì)象,但子元素也存在mutable可變對(duì)象(子元素部分mutable)
In [1]: a = [1, 2, ['hello','world']]

In [2]: import copy

In [3]: b = copy.deepcopy(a)

In [4]: id(a)
Out[4]: 140531593252104

In [5]: id(b)
Out[5]: 140531593479304

In [6]: [ id(item) for item in a ]
Out[6]: [9164864, 9164896, 140531593299016]

In [7]: [ id(item) for item in b ]
Out[7]: [9164864, 9164896, 140531593324232]

In [8]: [ id(item) for item in a[2] ]
Out[8]: [140531582302896, 140531582302784]

In [9]: [ id(item) for item in b[2] ]
Out[9]: [140531582302896, 140531582302784]

【解釋】:深拷貝既拷貝了頂層對(duì)象,又遞歸拷貝了子元素對(duì)象,所以a[2]與b[2]指向了兩個(gè)不同的列表對(duì)象(但是列表對(duì)象的子元素初始指定的字符串對(duì)象一樣),修改a[2][1] = 'china'后,它重新指向了新的字符串對(duì)象(內(nèi)存地址為140531581905808),不會(huì)影響到b[2][1]


  • (3)當(dāng)頂層對(duì)象是immutable不可變對(duì)象,同時(shí)它的子元素對(duì)象也全都是immutable不可變對(duì)象
In [1]: a = (1, 2, 3)

In [2]: import copy

In [3]: b = copy.deepcopy(a)

In [4]: id(a)
Out[4]: 140021832303960

In [5]: id(b)
Out[5]: 140021832303960

In [6]: [ id(item) for item in a ]
Out[6]: [9164864, 9164896, 9164928]

In [7]: [ id(item) for item in b ]
Out[7]: [9164864, 9164896, 9164928]

【嗯,對(duì)!】:變量a與變量b指向的是同一個(gè)元組對(duì)象,不存在拷貝


  • (4)當(dāng)頂層對(duì)象是immutable不可變對(duì)象時(shí),但子元素存在mutable可變對(duì)象(子元素部分mutable)
In [11]: a
Out[11]: (1, 2, ['hello', 'china'])

In [12]: b
Out[12]: (1, 2, ['hello', 'world'])

In [13]: [ id(item) for item in a[2] ]
Out[13]: [140437024839640, 140437016189336]

In [14]: [ id(item) for item in b[2] ]
Out[14]: [140437024839640, 140437024839528]

【不想多打字了...】:變量a與變量b指向的是不同的元組對(duì)象,同時(shí)a[2]與b[2]指向不同的列表對(duì)象,所以修改a[2][1]不會(huì)影響b[2][1]


其它拷貝方法

(1)列表的復(fù)制
列表的復(fù)制,有以下三種方式:

  • 列表的copy()函數(shù)

  • list()轉(zhuǎn)換函數(shù)

  • 列表分片[:]

In [1]: a = [1, 2, ['hello','world']]

In [2]: b = a.copy()

In [3]: c = list(a)

In [4]: d = a[:]

In [5]: id(a), id(b), id(c), id(d)
Out[5]: (140277244933640, 140277244846856, 140277323038536, 140277244767944)

In [6]: a[0] = 100

In [7]: a[2][1] = 'wangy'

In [8]: a
Out[8]: [100, 2, ['hello', 'wangy']]

In [9]: b
Out[9]: [1, 2, ['hello', 'wangy']]

In [10]: c
Out[10]: [1, 2, ['hello', 'wangy']]

In [11]: d
Out[11]: [1, 2, ['hello', 'wangy']]

【列表的復(fù)制都相當(dāng)于淺拷貝效果】
【解釋】:b/c/d都是a的復(fù)制,它們都指向了不同的列表對(duì)象,但是沒有拷貝子元素,a[2]和b[2]/c[2]/d[2]指向同一個(gè)列表,相當(dāng)于淺拷貝的效果


(2)元組的復(fù)制

In [1]: a = (1, 2, ['hello','world'])

In [2]: b = a[:]

In [3]: id(a), id(b)
Out[3]: (140146192445512, 140146192445512)

In [4]: a
Out[4]: (1, 2, ['hello', 'world'])

In [5]: b
Out[5]: (1, 2, ['hello', 'world'])

In [6]: a[2][1] = 'wangy'

In [7]: a
Out[7]: (1, 2, ['hello', 'wangy'])

In [8]: b
Out[8]: (1, 2, ['hello', 'wangy'])

【解釋】:使用分片[:]操作,a和b其實(shí)是指向同一個(gè)元組,而且沒有拷貝子元素,a[2]和b[2]也指向同一個(gè)列表,相當(dāng)于淺拷貝的效果


(3)字典的復(fù)制

In [1]: a = {'name': 'wangy', 'age': 18, 'jobs': ['devops', 'dba']}

In [2]: b = a.copy()

In [3]: c = dict(a)

In [4]: id(a), id(b), id(c)
Out[4]: (139653533041504, 139653544192616, 139653533040712)

In [5]: a['age'] = 20

In [6]: a['jobs'].append('python')

In [7]: a
Out[7]: {'name': 'wangy', 'age': 20, 'jobs': ['devops', 'dba', 'python']}

In [8]: b
Out[8]: {'name': 'wangy', 'age': 18, 'jobs': ['devops', 'dba', 'python']}

In [9]: c
Out[9]: {'name': 'wangy', 'age': 18, 'jobs': ['devops', 'dba', 'python']}

【解釋】:變量a與變量b/c指向不同的字典,但是沒有拷貝子元素,a['jobs']和b['jobs']/c['jobs']指定同一個(gè)列表,相當(dāng)于淺拷貝的效果


(4)集合的復(fù)制

同列表類似,可以使用集合的copy()函數(shù)或者轉(zhuǎn)換函數(shù)set()

In [1]: a = {1, 2, 3}

In [2]: b = a.copy()

In [3]: c = set(a)

In [4]: id(a), id(b), id(c)
Out[4]: (139965317888712, 139965317888936, 139965317889608)

In [5]: a.add('wangy')

In [6]: a
Out[6]: {1, 2, 3, 'wangy'}

In [7]: b
Out[7]: {1, 2, 3}

In [8]: c
Out[8]: {1, 2, 3}

【解釋】:變量a與變量b/c指向不同的集合,而集合的元素必須是hashable,所以修改集合a不會(huì)影響到b/c

最后編輯于
?著作權(quán)歸作者所有,轉(zhuǎn)載或內(nèi)容合作請(qǐng)聯(lián)系作者
平臺(tái)聲明:文章內(nèi)容(如有圖片或視頻亦包括在內(nèi))由作者上傳并發(fā)布,文章內(nèi)容僅代表作者本人觀點(diǎn),簡(jiǎn)書系信息發(fā)布平臺(tái),僅提供信息存儲(chǔ)服務(wù)。

推薦閱讀更多精彩內(nèi)容