Let's say I have something like this:
.
├── run.py
└── test
├── __init__.py
├── models
│ ├── foo
│ │ ├── baby.py
│ │ └── __init__.py
│ ├── __init__.py
│ └── user.py
└── start.py
run.py
from test import start
start.py
from .models import user
user.py
from . import foo
print(foo.baby.Baby)
baby.py
Baby = "I am a baby"
Now, when you run the run.py file...
>>> python run.py
... traceback ...
AttributeError: 'module' object has no attribute 'baby'
But, if you change the start.py like this:
from .models.foo import baby
from .models import user
everything works correctly.
When the baby module in start.py wasn't loaded earlier, the foo package object did not have a reference to it (foo.baby.Baby threw an AttributeError). But when I loaded the baby module in start.py, the foo package object automatically got a reference to babymodule.
Can someone explain me how this works?
Submodules are not automatically attributes of a package until imported.
You need to import test.models.foo.baby first before foo.baby.Baby works. You can do this in the foo/__init__.py file:
from . import baby
来源：https://stackoverflow.com/questions/16201068/a-python-module-and-package-loading-confusion