[ros-users] Experimental Results Repository

[Eitan Marder-Eppstein]

Ivan,

rosinstall will handle git, hg, bzr, and svn. See
http://www.ros.org/wiki/rosinstall#Version_Control_Modes for details. Though
the three DVCS's are marked as experimental, we've done a fair amount of
testing with them and things should work.

So, it should be quite easy to tell it to pull from your own git repository.

Let us know if you have any issues.

Hope all is well,

Eitan

On Wed, Sep 29, 2010 at 9:19 AM, Ivan Dryanovski
<ivan.dryanovski at gmail.com>wrote:

> Brian,
>
> Thank you for the reply. We have looked at the papers section and
> we'll try to structure our release in a similar way. One issue we are
> facing at the moment is what to do with downloading the code needed to
> reproduce the experiment. We saw the ROS papers use a .rosconfig file
> and pass that to rosinstall, in order to pull all the packages at the
> correct revision. Our code, however, relies on both the ROS repository
> (svn) and our own repo (git).
>
> Does the rosinstall tool handle git version control? If not, what
> would you recommend we do for now? I am thinking about making a branch
> of our repo (lets say, ICRA2011), and adding manual installation
> instructions in the paper wiki. So something like:
>
> cd ~/ros/stacks
> git clone http://robotics.ccny.cuny.edu/git/ccny-ros-pkg.git
> cd ~/ros/stacks/ccny-ros-pkg
> git checkout ICRA2011
> rosmake foo ... etc
>
> Thanks,
>
> Ivan Dryanovski
>
> On Wed, Sep 29, 2010 at 1:14 AM, Brian Gerkey <gerkey at willowgarage.com>
> wrote:
> > hi Bill,
> >
> > First off, let me say that we wholeheartedly support your desire to
> > enable reproduction of your experiments.  It's a practice from which
> > the community will benefit greatly.
> >
> > We've been doing some work in this vein, which you can see in the wiki:
> >  http://www.ros.org/wiki/Papers
> > The idea is to create a page for each paper (unpublished experiments
> > could also go there), with enough information to run the experiment
> > yourself.  In an ideal world, the reader would be able to recreate the
> > graphs or tables that appear in the paper.  A good example is this
> > ICRA 2010 paper:
> >  http://www.ros.org/wiki/Papers/ICRA2010_Marder-Eppstein
> > Toward the bottom are instructions for getting and compiling the code,
> > and running the experiments that are reported on in the paper, both in
> > simulation and on the robot (if you happen to have a robot of the sort
> > that was used).
> >
> > NOTE: The Papers page in the ros.org wiki is open to everybody.   We
> > encourage other ROS users to put up pages describing their papers.
> >
> > Now, to your question of where to keep the code: so far, we've been
> > putting packages dedicated our experimental results in our sandbox.
> > E.g., the ICRA paper I mentioned relies on a package called
> > 'icra_navigation_gazebo', which does get indexed into
> > http://www.ros.org/browse/list.php .
> >
> > As you point out, it might be good to filter this sort of package out
> > of the rosbrowse list.  Perhaps we could add support for a
> > robots.txt-like file that maintainers can drop in parts of their repos
> > to tell rosbrowse to go away.
> >
> >        brian.
> >
> > On Mon, Sep 27, 2010 at 10:59 AM, Bill Morris <morris at ee.ccny.cuny.edu>
> wrote:
> >> Here at the CCNY Robotics Lab we are looking at archiving our
> >> experimental results. We want to be able to provide enough information
> >> so that our results can be reproduced easily.
> >>
> >> Currently we are debating the creation of a separate repository
> >> ccny-ros-exp which would contain installers and launch files necessary
> >> to rerun our experiments. This approach in nice in that it keeps from
> >> cluttering up our main repository and most users of our code won't
> >> require our experimental data. The downside is that it clutters up the
> >> ROS list of repositories.
> >>
> >> Perhaps there should be a separate page like this
> >> http://www.ros.org/browse/repo_list.php just for experimental results.
> >> This could also be indexed so that you could look at a sensor message
> >> and have a list of experiments that used that message. I'm sure there
> >> are other benefits that could be found.
> >>
> >> Does anyone have any opinions or feedback on creating a separate
> >> repository for experimental data?
> >>
> >> _______________________________________________
> >> ros-users mailing list
> >> ros-users at code.ros.org
> >> https://code.ros.org/mailman/listinfo/ros-users
> >>
> > _______________________________________________
> > ros-users mailing list
> > ros-users at code.ros.org
> > https://code.ros.org/mailman/listinfo/ros-users
> >
> _______________________________________________
> ros-users mailing list
> ros-users at code.ros.org
> https://code.ros.org/mailman/listinfo/ros-users
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.ros.org/pipermail/ros-users/attachments/20100929/01289ec4/attachment-0003.html>