+1

I wonder if this could also be used to control which version of Python you are using without changing the paths and such.  I have been working on making ros, ros_comm, and common_msgs more Python 2/3 portable because of MORSE, which uses Blender, which requires Python3.  I had considered setting something like env['ROS_PYTHON'] and replacing the #!/usr/bin/env python lines with something like ros-shebang that would select python executables and paths based on the environment variable.  This isn't something I see people doing often, but it might be a use case to consider when designing this tool.  There might also be a better/existing way of doing what I am talking about, but I am just not aware of it.

Looks good,

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
William Woodall
Graduate Software Engineering
Auburn University
w@auburn.edu
wjwwood@gmail.com
williamjwoodall.com
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~



On Wed, Aug 10, 2011 at 11:21 AM, Ken Conley <kwc@willowgarage.com> wrote:
+1

If shorter names are being debated, I offer 'rosbang'.  The placement
of the hyphen in ros-shebang doesn't quite feel like a ros command
name.

I should also note that part of this discussion also includes what
distributions ros-shebang should be included in, i.e. should it be
backported to diamondback, sneak in through the electric freeze,
etc...

The hashbang spec has definitely been an impediment for having a good
ease-of-use with rosh, so I support at least having this in ROS
Electric.  I'm indifferent to ROS Diamondback as my development is
always focused on the latest-and-greatest release.

 - Ken

On Tue, Aug 9, 2011 at 2:40 PM, Thibault Kruse <kruset@in.tum.de> wrote:
> Hi,
>
> I would like to make a suggestion.
>
> I want to add a command (called ros-shebang for now) to ros/bin that runs
> script files by invoking a second shebang line.
>
> This allows invoking script interpreters with #! and passing additional
> arguments, as well as invoking interpreters that live in ROS packages.
>
> My stake is calling an interpreter for roslisp scripting. However a
> similar concern is invoking of rosh plugins, see
> http://www.ros.org/wiki/rosh/Overview/Roshlets
>
> Example for rosh:
> #! /usr/bin/env ros-shebang
> #! /usr/bin/env rosrun rosh rosh rosh/echolet.py
> --plugins=rosh_common,rosh_geometry
>
> Example for roslisp:
>
> #! /usr/bin/env ros-shebang
> ;; /usr/bin/env rosrun sbcl run-sbcl.sh --script
>
> One less known alternative is to use other directives to declare the
> script interpreter than #!, (in case you wonder, for lisp, #| and ":";
> work) however this does not work with python subprocess.Popen, which means
> roslaunch and rostest would fail to run scripts like that (except using
> shell=True, but this is strongly discouraged for security and portability
> reasons).
>
> Alternatives:
> - A modification to roslaunch syntax to allow forcing shell=true
>
> - an executable in ros/bin that executes roslisp scripts, such that we
> could write:
> #! /usr/bin/env ros-run-roslisp-script
>
> - A required fixed installation of our lisp implementation, or some custom
> made variant of /usr/bin/env
>
>
> I wrote a ros-shebang script in python that works, which is no more magic
> than reading the second line and calling python execvp on it (minus the
> first two symbols).
>
> If a new command were introduced to ros/bin, this would affect more people
> than those working with roslisp, so opinions should be heard. There is no
> urgency to this, roslisp works fine as it is without offering comfortable
> scripting, just as rosh works fine by loading plugins after the shebang
> line.
>
> Not sure if there are other scripting interpreter usages with similar
> issues in ROS (Java, Lua, Javascript, ...).
>
> Thoughts?
>
>  Thibault
>
> _______________________________________________
> ros-developers mailing list
> ros-developers@code.ros.org
> https://code.ros.org/mailman/listinfo/ros-developers
> _______________________________________________
> ros-users mailing list
> ros-users@code.ros.org
> https://code.ros.org/mailman/listinfo/ros-users
>
_______________________________________________
ros-users mailing list
ros-users@code.ros.org
https://code.ros.org/mailman/listinfo/ros-users