Why do Python modules sometimes not import their sub-modules?

Question

I noticed something weird today I would like explained. I wasn't 100% sure how to even phrase this as a question, so google is out of the question. The logging module does not have access to the module logging.handlers for some odd reason. Try it yourself if you don't believe me:

>>> import logging
>>> logging.handlers
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
AttributeError: 'module' object has no attribute 'handlers'
>>> import logging.handlers
>>> logging.handlers
<module 'logging.handlers' from '/usr/lib/python2.6/logging/handlers.pyc'>

Can anyone explain why this happens?

Answer 1

In Python, modules need to be imported before they're accessible. import logging imports just the logging module. It so happens that logging is a package with submodules, but those submodules are still not automatically loaded. So, you need to explicitly import logging.handlers before you can access it.

If you're wondering why it looks like sometimes you don't need those extra imports: some packages import some or all of their submodules when they are imported -- simply by doing those imports in their __init__.py files. In other cases it might be that something else that you import, also imported logging.handlers. It doesn't matter which piece of code does the import; as long as something in your process imports logging.handlers before you access it, it'll be there. And sometimes a module that looks like a package really isn't one, like os and os.path. os isn't a package, it just imports the correct other module (for your platform) and calls it path, just so you can access it as os.path.