无法从子目录导入模块

Question

On a high level the structure of my project looks like this.

dir/
    moduleA/
            __init__.py
            file1.py
            file2.py
    main.py

file2.py contains some global variable that I am importing in file1.py

# --- file2.py schema ---

SOME_GLOBAL_VARIABLE = 1

.

# --- file1.py schema --- 

from moduleA.file2 import SOME_GLOBAL_VARIABLE

class MyClass():
    def __init__(self):
         self.some_attr = SOME_GLOBAL_VARIABLE   

    def run():
         print(self.some_attr)     

When I run main.py below,which creates an instance of MyClass, the code runs without errors.

# --- main.py ---

from moduleA.file1 import MyClass

x = MyClass()
x.run()

However, when I open a open a Python console and execute the below to import MyClass from file1.py I get an import error that there is no module named moduleA.

from moduleA.file1 import MyClass

# (... some traceback statements)
from moduleA.file2 import SOME_GLOBAL_VARIABLE
ModuleNotFoundError: No module named 'moduleA'

How is this possible when Python clearly finds moduleA/file1.py, otherwise it could not try to execute the line

from moduleA.file2 import SOME_GLOBAL_VARIABLE

What seems strange to me is that it runs without errors when I run the main.py script.

I guess this is due to my lack in understanding how Python exactly imports modules, so any help is greatly appreciated. I tried adding the init .py file in moduleA but this did not resolve the issue.

In case this information is of relevance, the path some_path/dir is on my PYTHONPATH.

Answer 1

This is caused because of the __package__ variable.
To see it's value, add print(__package__) in file1.py before the import.

When running file1.py directly, you can see that __package__ is None .
But on the other hand, when running main.py , when importing file1.py (and thus running it), __package__ 's value (in file1.py ) is suddenly moduleA , which is why python is able to import moduleA.file2 .