It's All Geek to Me

Misleading ImportErrors in Django

I was debugging an issue with our Django app at work today; an admin.py file wasn't being picked up, and nothing was appearing in the admin pages. It turned that an ImportError was being thrown in the admin.py and Django was interpreting this as the file not existing.

I'm assuming that the reason for this is that Django uses __import__ to import the module, and catches ImportError's if the admin.py doesn't exist. The trouble with this is that if admin.py does exist, and throws an ImportError of its own, Django will also interpret that as a missing admin.py – which can be misleading.

The only solution I can think of to more accurately handle this would be to programmaticaly examine the traceback to determine where the ImportError is thrown. If the traceback is one level deep, we know that the ImportError was thrown because the module doesn't exists. If it is greater than one level then we know the module was found, but has thrown an ImportError of its own.

Here's a simple proof of concept:

#check_import.py
import traceback
import sys

def check_import(name):
    try:
        __import__(name)
    except ImportError:
        exceptionType, exceptionValue, exceptionTraceback = sys.exc_info()
        return len(traceback.extract_tb(exceptionTraceback)) > 1
    return True

print check_import("os")
print check_import("noimport")
print check_import("anotherlevel")

#anotherlevel.py
import thismoduledoesnotexist

This produces the following when run:

True
False
True

It would be nice if Django did something similar for its implicit imports. I think the best behaviour would be re-raise the ImportError if it the module does actually exist. That way, it is clear what the problem is. I may attempt to write a patch at some point, unless someone knows of a better solution.