I’m trying to understand how pickling works so that I can correctly design the organization of my repository. Essentially, I pickle objects of type A, whose instances contain objects of type B. A and B are defined in modules in separate packages:

project_dir
|
|-- package_a
  |
  |-- module_a  # class A is defined here
|
|-- package_b
  |
  |-- module_b # class B is defined here

I did an experiment that I expected to fail, given my knowledge about how pickling works, but it didn’t throw an exception:

module_a.py:

class A:
    name = "A"

module_b.py:

class B:
    def __init__(self) -> None:
        self.a = None

In one terminal window:

project_dir> python
>>> import pickle
>>> from package_a.module_a import A 
>>> from package_b.module_b import B 
>>> b_obj = B()   
>>> b_obj.a = A() 
>>> b_obj.a.name  
'A'
>>> with open("b.pickle", "wb") as f:
...     pickle.dump(b_obj, f)
...
>>> exit()

In a new terminal window:

project_dir> python
>>> import pickle
>>> from package_b.module_b import B
>>> with open("b.pickle", "rb") as f:
...     b_obj = pickle.load(f)
... 
>>> b_obj.a.name 
'A'

How is pickle able to construct an A object when I haven’t imported module_a anywhere?