I want to distribute my app on OSX (using py2app) and as a Debian package.
The structure of my app is like:
app/
     debian/
            <lots of debian related stuff>
     scripts/
             app
     app/
         __init__.py
         app.py
         mod1/
              __init__.py
              a.py
         mod2/
              __init__.py
              b.py
My setup.py looks something like:
from setuptools import setup
import os
import os.path
osname = os.uname()[0]
if osname == 'Darwin':
    APP = ['app/app.py']
    DATA_FILES = []
    OPTIONS = {'argv_emulation': True}
    setup(
        app=APP,
        data_files=DATA_FILES,
        options={'py2app': OPTIONS},
        setup_requires=['py2app'],
    )
elif osname == 'Linux':
        setup(
        name = "app",
        version = "0.0.1",
        description = "foo bar",
        packages = ["app", "app.mod1", "app.mod2"],
        scripts = ["scripts/app"],
        data_files = [
            ("/usr/bin", ["scripts/app"]),
       ]
    )
Then, in b.py (this is on OSX):
from app.mod2.b import *
I get:
ImportError: No module named mod2.b
So basically, mod2 can't acccess mod1. On Linux there's no problem, because the python module 'app' is installed globally in /usr/shared/pyshared. But on OSX the app will obviously be a self-contained .app thing built by py2app. I wonder if I approached this totally wrong, are there any best practices when distributing Python apps on OSX?
Edit: I also tried a hack like this in b.py:
from ..mod2.b import *
ValueError: Attempted relative import beyond toplevel package
Edit2: Seems to be related to this How to do relative imports in Python?
 
     
    