Dict kwargs recursive wrapping breaks reference assignment about addict

[pohmelie]

Also, this creates disambiguations:

>>> import addict
>>> addict.__version__
'2.1.1'
>>> from addict import Dict
>>> l = [1, 2, 3]
>>> d = Dict(l=l)
>>> d.l
[1, 2, 3]
>>> d.l is l
False
>>> d.l = l
>>> d.l is l
True
>>>

[egao1980]

This is more of documentation issue than a bug.
Dict() copies dict/list/tuple structure so this is expected behaviour.

    def test_recursive_wrapping_refs(self):
      d = Dict(a=1, b=2)
      a = Dict(d=d)
      c = Dict(d=d)
      c.d.z = 3
      self.assertDictEqual(a.d, d)
      self.assertDictContainsSubset(d, c.d)
      self.assertIsNot(a.d, d)
      self.assertIsNot(c.d, d)

[pohmelie]

It is implicit copying. Don't you think explicit using of copy module will be more pythonic, so Dict behavior will be closer to dict.

[mewwts]

Hey @pohmelie,

I understand that the behavior is a bit confusing. It's an attempt to balance between expectations of the addict module. Today, whatever you give addict in the constructor will be recursively iterated over and any dict will be turned into a Dict. In contrast, whatever you set as an attribute on the Dict instance will not be copied. While this is the way we intended it to work, it can lead to surprises.

At this point, I am leaning towards moving today's __init__-functionality in it's own static method, say Dict.from_dict, and let __init__ do as you expect over. This would however warrant a major release.

Thanks @egao1980 for stepping in and answering questions! 👍

[hellocoldworld]

hello @mewwts is the new from_dict method already available? I think it could greatly improve the reading of the code, so if it's not done, I think it would be a great idea!