Import redirects

Isaac To isaac.to at gmail.com
Mon Feb 11 07:50:36 CET 2013


I have a package (say "foo") that I want to rename (say, to "bar"), and for
compatibility reasons I want to be able to use the old package name to
refer to the new package.  Copying files or using filesystem symlinks is
probably not the way to go, since that means any object in the modules of
the package would be duplicated, changing one will not cause the other to
be updated.  Instead, I tried the following as the content of
`foo/__init__.py`:

    import sys
    import bar
    sys.modules['foo'] = bar

To my surprise, it seems to work.  If I `import foo` now, the above will
cause "bar" to be loaded and be used, which is expected.  But even if I
`import foo.baz` now (without first `import foo`), it will now correctly
import "bar.baz" in its place.

Except one thing: it doesn't really work.  If I `import foo.baz.mymod` now,
and if in "bar.baz.mymod" there is a statement `import bar.baz.depmod`,
then it fails.  It correctly load the file "bar/baz/depmod.py", and it
assigns the resulting module to the package object bar.baz as the "depmod"
variable.  But it fails to assign the module object of "mymod" into the
"bar.baz" module.  So after `import foo.baz.mymod`, `foo.baz.mymod` results
in an AttributeError saying 'module' object has no attribute 'mymod'.  The
natural `import bar.baz.mymod` is not affected.

I tested it with both Python 2.7 and Python 3.2, and I see exactly the same
behavior.

Anyone knows why that happen?  My current work-around is to use the above
code only for modules and not for packages, which is suboptimal.  Anyone
knows a better work-around?
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/python-list/attachments/20130211/5073ad86/attachment.html>


More information about the Python-list mailing list