19
votes

I know that there are lots of questions about the same import issues in Python but it seems that nobody managed to provide a clear example of correct usage.

Let's say that we have a package mypackage with two modules foo and bar. Inside foo we need to be able to access bar.

Because we are still developing it, mypackage is not in sys.path.

We want to be able to:

  • import mypackage.foo
  • run foo.py as a script and execute the sample usage or tests from the __main__ section.
  • use Python 2.5

How do we have to do the import in foo.py in order to be sure it will work in all these cases.

# mypackage/__init__.py
...

# mypackage/foo/__init__.py
...

# mypackage/bar.py  
def doBar()
    print("doBar")

# mypackage/foo/foo.py
import bar # fails with module not found
import .bar #fails due to ValueError: Attempted relative import in non-package

def doFoo():
    print(doBar())

if __name__ == '__main__':
    doFoo()
2
I believe you have to distinguish importable modules from scripts. I'm not sure I agree with that design, but that's my understanding of it.Thomas K
Usually, when you need to import packages from upper level, you are doing something wrong: would you instead want to use bar as a library rather than as a submodule to mypackage? I.e. could you refactor mypackage into mypackage1 and mypackage2 where mypackage2 (with foo) imports mypackage1 (with bar)?jsalonen

2 Answers

32
votes

Take a look at the following info from PEP 328:

Relative imports use a module's __name__ attribute to determine that module's position in the package hierarchy. If the module's name does not contain any package information (e.g. it is set to '__main__') then relative imports are resolved as if the module were a top level module, regardless of where the module is actually located on the file system.

When you run foo.py as a script, that module's __name__ is '__main__', so you cannot do relative imports. This would be true even if mypackage was on sys.path. Basically, you can only do relative imports from a module if that module was imported.

Here are a couple of options for working around this:

1) In foo.py, check if __name__ == '__main__' and conditionally add mypackage to sys.path:

if __name__ == '__main__':
    import os, sys
    # get an absolute path to the directory that contains mypackage
    foo_dir = os.path.dirname(os.path.join(os.getcwd(), __file__))
    sys.path.append(os.path.normpath(os.path.join(foo_dir, '..', '..')))
    from mypackage import bar
else:
    from .. import bar

2) Always import bar using from mypackage import bar, and execute foo.py in such a way that mypackage is visible automatically:

$ cd <path containing mypackage>
$ python -m mypackage.foo.foo
7
votes

My solution looks a bit cleaner and can go at the top, with all the other imports:

try:
   from foo import FooClass
except ModuleNotFoundError:
   from .foo import FooClass