从一个子模块导入到另一个子模块

Question

I'm struggling with python's imports.

I have my repository for my python projects in which I have a collection of more or less independent programmes and tools. There are a few sub-folders which contain the single scripts and programmes; some can be used as helper routines for others scripts as well. Assume that my folder structure looks like:

root
   __init__.py
   prog1.py
   prog2.py
   imaging
        __init__.py
        filter1.py
        detect_edges.py
   data_tools
        __init__.py
        floating_mean.py
        unify_data_sources.py
        prog3.py

Now assume that I want to use prog3.py directly but need call filter1.py . How can I import it into prog3.py ?

The only way I managed to import it was giving adding to the path which needs modification of sys.path .

import sys
sys.path.append("../imaging/")
from filter1 import filter1

As I know the relative way the modules are located, is there another way similar to what is described in python's module import guide ( here and here ) without changing sys.path?

Answer 1

Assuming you need to call filter1.py for importing shared functions or classes, I would use relative imports as described in the python tutorial.

filter1.py

def some_shared_func1():
    pass

def some_shared_func2(): 
    pass

prog3.py

from ..imaging.filter1 import some_shared_func1, some_shared_func2
# filter1 is reached from prog3 by go two levels up then one level down to `imaging` folder.

if __name__ == "__main__":
    # Do something with some_shared_func1
    # Do something with some_shared_func2
    pass

I would then run prog3.py with -m argument to resolve relative imports but without extension:

python -m root.data_tools.prog3

Current directory must be the one containing root (or be defined into PYTHONPATH variable).

I don't modify sys.path .

Is that what you are looking for?