Re: [Jython-dev] difference in semantics for handling parallel directories in Java and Python

[Frank W. <fwi...@gm...>]

On 7/21/07, Charlie Groves <cha...@gm...> wrote:
> This is different than what sys.path does, since multiple python
> package directories will be merged together by the classloader.
> However, I don't see this as that big of an issue since sys.path is
> still there if you want the Python way of doing things.
That's fair.  Those wanting/expecting python semantics should use
python.path, those who want Java semantics can put the py files on the
CLASSPATH.  I wonder if we should pass some system variables to
python.path in the generated shell and .bat files by default?  We
could use $JYTHONPATH as a sort of parallel to $PYTHONPATH (on windows
I guess it would be %JYTHONPATH%) and then document it -- that way it
will be closer to what CPython developers would expect.  I am getting
ready to check in a shell and a .bat file for developer convenience
into the repo so we can experiment with these sorts of ideas -- I'm
modeling them after the shell/.bat files you can find with tomcat or
ant where the script checks for cygwin, etc).  The biggest remaining
problem I see is: which style should win when a path is on both the
CLASSPATH and python.path?

-Frank