[Idle-dev] Re: idlefork merge underway - errata

[Guido van Rossum]

> Guido van Rossum wrote:
> > > Guido, I just merged idle.py, and am moving on to it's bretheren, but I'm
> > > curious about the change:
> > >
> > > ! idle_dir = os.path.split(sys.argv[0])[0]
> > > ---
> > > ! idle_dir = os.path.dirname(IdleConf.__file__)
> > >
> > > What is the reason for moving to the second form of working out idle_dir
> > > ??
> >
> > This is trying to find IDLE's source directory so that it can look
> > there for config files and icons (I think).
> >
> > The old form just *assumes* that sys.argv[0] is the absolute pathname
> > of idle.py.  This isn't always the case (depends on shell and OS).
> >
> > The second one assumes that the IdleConf module lives in the IDLE
> > source directory.  That's a pretty good assumption I believe!
> 
> Isn't   
> 
> idle_dir = os.path.abspath(sys.path[0]) + os.sep   
> 
> os independent, and without worrying about the existence of particular files 
> or modules??

Concatenating os.sep is not OS dependent.  On the Mac, os.sep is ':'
but '::' is the equivalent of '../' and ':::' is '../../', etc.  I've
also seen Windows balk about excess trailing slashes in some cases.

On Unix, sys.path[0] is not the right place to look into.  It might be
just "idle" when the "idle" script is found on $PATH and the user
typed "idle".  (Most shells sustitute the full pathname found, but not
all do.)

So please just do what I said. :-)

--Guido van Rossum (home page: http://www.python.org/~guido/)