python: Why listA.append('a') affects listB?

Question

This is my code:

In [8]: b=dict.fromkeys([1,2,3,4], [])

In [9]: b[1].append(1)

In [10]: b[2].append(2)

In [11]: b[1]
Out[11]: [1, 2]

In [12]: b[2]
Out[12]: [1, 2]

In [13]: b
Out[13]: {1: [1, 2], 2: [1, 2], 3: [1, 2], 4: [1, 2]}

Whereas I would expect: {1: [1], 2: [2], 3: [], 4: []}

I guess this is maybe caused by b[X] is just a "reference", they all refer to a same list.

Then I replace [] with an int object. The result makes me more confused:

In [15]: b=dict.fromkeys([1,2,3,4], 1)

In [16]: b[1] += 1

In [17]: b[2] += 1

In [18]: b
Out[18]: {1: 2, 2: 2, 3: 1, 4: 1}

This int object 1 isn't a referance in this case.

Then I replace [] with ['a']:

In [19]: b=dict.fromkeys([1,2,3,4], ['a'])

In [20]: b[1].append(1)

In [21]: b[2].append(2)

In [22]: b
Out[22]: {1: ['a', 1, 2], 2: ['a', 1, 2], 3: ['a', 1, 2], 4: ['a', 1, 2]}

Now ['a'] is a reference again.

Can someone tells me why, and how to get expected result "{1: [1], 2: [2], 3: [], 4: []}" in the first case.

Any useful suggestions are appreciated.

Ashwini Chaudhary · Accepted Answer · 2013-10-03T17:41:26.690

6

Because all values in the dict are actually references to the same list, dict.fromkeys uses the same list object and assigns it to each key. As list.append is an in-place operation so all keys are affected.

>>> b = dict.fromkeys([1,2,3,4], [])
>>> [id(x) for x in b.values()]
[158948300, 158948300, 158948300, 158948300]

So, for mutable value use a dict comprehension:

>>> b = {k:[] for k in xrange(1, 5)}
>>> [id(x) for x in b.values()]
[158945580, 158948396, 158948108, 158946764]

Or as @Bakuriu suggested, collections.defaultdict will also work fine:

>>> from collections import defaultdict
>>> dic = defaultdict(list)
>>> dic[1].append(1)
>>> dic[2].append(2)
>>> dic
defaultdict(<type 'list'>, {1: [1], 2: [2]})
>>> dic[3]
[]

edited Oct 03 '13 at 17:41

answered Oct 03 '13 at 17:35

Ashwini Chaudhary

244,495
58
464
504

2

Or even better: use a `defaultdict` which avoids the need to use `fromkeys` altogether. – Bakuriu Oct 03 '13 at 17:36
Thanks, it makes perfect sense now. – Oct 04 '13 at 02:00

Mark Ransom · Answer 2 · 2013-10-03T18:02:53.553

4

They really are all references. The difference is what you do with the references.

When you use the assignment operator =, you're setting the reference to a different object. (+= works the same).

When you use append, you're modifying the object without affecting the reference. Since fromkeys gave back multiple references to the same object, the modifications are seen in all of them simultaneously.

edited Oct 03 '13 at 18:02

answered Oct 03 '13 at 17:42

Mark Ransom

299,747
42
398
622

This conceptual explanation actually does address the OP's question, +1 – iruvar Oct 03 '13 at 17:51
Thanks, this answer also added to my understanding. – Oct 06 '13 at 02:50

python: Why listA.append('a') affects listB?

2 Answers2

Linked