Dataclass nested within another dataclass does not update data correctly

Question

I generate two different instances of a python dataclass which includes a nested dataclass. When I update a value in the nested dataclass in one instance (but not in the other), the same data is placed in the nested dataclass in both instances. This is not what I expected.

from dataclasses import dataclass


@dataclass
class sub1:
    q: int = 10
    r: str = "qrst"


@dataclass
class A:
    a: int = 1
    s1: sub1 = sub1()


if __name__ == '__main__':
    a = A()
    aa = A()
    aa.a = 9
    aa.s1.r = "92"
    print("a:", repr(a))
    print("aa:", repr(aa))

''' Produces --
a: A(a=1, s1=sub1(q=10, r='92'))
aa: A(a=9, s1=sub1(q=10, r='92'))
'''

I expected the nested dataclass to be updated in only the specified instance (aa) and for the nested dataclass in the other instance (a) to remain unchanged.

What am I doing wrong, or is dataclass the wrong tool?

This has nothing to do with dataclasses, this would work the same way if you did `s1: sub1 = []` This is how default values *always work*. — juanpa.arrivillaga, Sep 26 '19 at 23:27
Also note, this isn't "nesting". You are simply using an object as an attribute of another object, *composition* would be the jargon. Also, please always use the generic [python] tag for all python related questions. — juanpa.arrivillaga, Sep 26 '19 at 23:27
This is essentially a duplicate of this: https://stackoverflow.com/questions/1680528/how-to-avoid-having-class-data-shared-among-instances although, perhaps the involvement of data-classes makes it worthy of it's own question, but the answer/root cause is essentially the same. — juanpa.arrivillaga, Sep 26 '19 at 23:33
This is not a generic python question as it is python 3 that has dataclasses. — Den, Sep 28 '19 at 02:17
it *is a Python question* so it should be tagged with the generic tag. Use a version-specific tag at your discretion — juanpa.arrivillaga, Sep 28 '19 at 02:19
juanpa. I've thought about this a bit more, and you're right. I'll be careful about this in the future. — Den, Sep 30 '19 at 01:16

score 12 · Answer 1 · answered Sep 26 '19 at 23:54

What you are currently doing is providing a default value for the field. As that value is a mutable object, changes to that object will be visible to all instances of your dataclass.

What you should do instead is provide a default factory that produces sub1 instances for each new A instance:

from dataclasses import field

@dataclass
class A:
    a: int = 1
    s1: sub1 = field(default_factory=sub1)

a = A()
aa = A()
aa.a = 9
aa.s1.r = "92"
print("a:", repr(a))  # a: A(a=1, s1=sub1(q=10, r='qrst'))
print("aa:", repr(aa))  # aa: A(a=9, s1=sub1(q=10, r='92'))

Dataclass nested within another dataclass does not update data correctly

1 Answers1