Pydantic 1.6: fields with default_factory cause parse_obj() to set fields to unparsed values

[johnanthonyowens]

Bug

This is likely related to #1710 but does not explicitly involve validators. This would be a good candidate for a unit test to prevent future regressions.

Output of python -c "import pydantic.utils; print(pydantic.utils.version_info())":

pydantic version: 1.6
pydantic compiled: True
install path: /usr/local/lib/python3.8/site-packages/pydantic
python version: 3.8.3 (default, May 16 2020, 07:08:28)  [GCC 8.3.0]
platform: Linux-4.15.0-109-generic-x86_64-with-glibc2.2.5
optional deps. installed: []

Here is a self-contained example illustrating the problem:

from pydantic import BaseModel, Field

class Inner(BaseModel):
    val: int = Field(0)

class Outer(BaseModel):
    inner_1: Inner = Field(default_factory=Inner)
    inner_2: Inner = Field(Inner())

default = Outer().dict() # default == {'inner_1': {'val': 0}, 'inner_2': {'val': 0}}
parsed = Outer.parse_obj(default)

print(parsed)

In 1.6, this prints:
inner_1={'value': 'default'} inner_2=Inner(value='default')
Note that inner_1 is a dict, not an Inner object!

In 1.5.1 this prints:
inner_1=Inner(val=0) inner_2=Inner(val=0)

[PrettyWood]

Hello @johnanthonyowens you are right !
I added your example in the unit tests of the fix 16e489b and updated the changelog.
Thanks for the suggestion as the regression did not only impact the validation of nested models but also the parsing!