PYTHONPATH environment variable…how do I make every subdirectory afterwards?

Question

I currently do this:

PYTHONPATH=/home/$USER:/home/$USER/respository:/home/$USER/repository/python-stuff

How can I make it so that the PYTHONPATH can include everything subdirectory?

PYTHONPATH = /home/$USER/....and-all-subdirectories

Answer 1

That's not how PYTHONPATH works; PYTHONPATH treats its search path differently from shell PATH. Let's say I do this:

$ mkdir /home/jsmith/python
$ cd /home/jsmith/python
$ touch a.py b.py

This will work, in Python (sys.path will include the current directory):

$ cd /
$ PYTHONPATH=/home/jsmith/python python2.6

Python 2.6.1 (r261:67515, Jul  7 2009, 23:51:51)
>>> import a, b   # Works
>>> quit()

However, subdirectories are treated as packages when __init__.py is present in the directory, and are ignored by PYTHONPATH otherwise:

$ mkdir /home/jsmith/python/pkg
$ cd /home/jsmith/python/pkg
$ touch __init__.py c.py d.py
$ cd /
$ PYTHONPATH=/home/jsmith/python python2.6

Python 2.6.1 (r261:67515, Jul  7 2009, 23:51:51)
>>> import a, b   # Works
>>> import c
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
ImportError: No module named c

To get at something in that subdirectory, this would work:

>>> from pkg import c   # Works
>>> import pkg.c        # Works

To roll a solution where every subdirectory in your PYTHONPATH is added, you need to explicitly add every folder to PYTHONPATH or sys.path programmatically. This behavior is intentional, and behaves nothing like shell PATH. Given the interpreter's support for packages in this regard, surely there's a better way to accomplish what you're after?

• Understanding imports and PYTHONPATH

Answer 2

That's not how environment PATH variables work - you give it the top-level directory and it's up to the application to recurse the directory tree if it needs to.