Question

The import function of Python still confuses me sometimes. Here’s an example:

My project has the following package structure:

Project/
    src/
        example/
            __init__.py
            an_example.py
        top/
            __init__.py
            lin/
                __init__.py
                factory.py

In an_example.py, I’d like to write

from top import lin
if __name__ == '__main__':
    a = lin.factory.AClass()

However, this fails with:

    a = lin.factory.AClass()
AttributeError: 'module' object has no attribute 'factory'

an_example.py works when written like this:

from top.lin import factory
if __name__ == '__main__':
    a = factory.AClass()

Can you explain to me why it is wrong to write the import statement like the first version? I’d prefer a fully qualified name like lin.factory.AClass to factory.AClass.

Answer 1

This is because, unless you tell lin to import factory in __init__.py, then factory is not in the lin namespace.

E.g:

Presume your existing project structure, with an_example.py containing:

from top import lin

lin.factory.AClass()

With top/lin/__init__.py blank, we get ImportError: No module named factory.

With top/lin/__init__.py containing import top.lin.factory, we get no error.

When you ask to use factory.AClass(), it works because it is defined there. Likewise, you need to define factory in lin if you want to use it from there.