[ros-users] [Ros-developers] Fwd: Review for Dynamic Reconfigure Groups

[Thibault Kruse]

Hi,

a bit late maybe to mention this, but I find it strange that 
dynamic_reconfigure
would invent its own type system. Why not build on the
ROS msg type system and work it from there? That would reduce
the duplicate code (binding generation) and allow for easier integration 
with other
programming languages. It does not even need to support the whole
range of ros msg possibilities for that to start with.

"Grouping parameters" sounds very much like building a ROS message 
structure to me.

I understand that you cannot define variable domains in ros msg definitions,
but why should a dynamic reconfigure file not start by loading a message
struct from a ros msg definition, and extend this to a configuration?

If you look at the example on using dynamic reconfigure merged with a 
service:
http://ibotics.ucsd.edu/trac/stingray/wiki/ROSNodeTutorialC%2B%2B
  this bit seems like redundant code:

void  NodeExample::messageCallback(const  node_example::node_example_data::ConstPtr&msg)
{
   message_=  msg->message;
   a_=  msg->a;
   b_=  msg->b;
...
}  // end publishCallback()
void  NodeExample::configCallback(node_example::node_example_paramsConfig&config,  uint32_t level)
{
   // Set class variables to new values. They should match what is input at the dynamic reconfigure GUI.
   message_=  config.message.c_str();
   a_=  config.a;
   b_=  config.b;
}  // end configCallback()

If dynamic reconfigure was based on ros msg, one callback would do for both.
That in itself is not a huge benefit to anyone maybe, it just would seem more
reasonable to me than to have two parallel type systems.


In the same way, i would have been happy to see dynamic_reconfigure cfg
files being yaml files rather than python, declaring structures only.
That could one day even lead to ROS msg definitions which include more
semantics like minimum, maximum, default, docstring. Is there a good
reason why this cannot or should not be done?

cheers,
   Thibault



On 01/-10/-28163 08:59 PM, Ken Conley wrote:
> Forwarding to ros-developers as well
>
>
> ---------- Forwarded message ----------
> From: Eitan Marder-Eppstein<eitan at willowgarage.com>
> Date: Thu, Jul 7, 2011 at 8:09 AM
> Subject: Review for Dynamic Reconfigure Groups
> To: ros-review at lists.willowgarage.com
> Cc: Ze'ev Klapow<klapow at willowgarage.com>, Ken Conley<kwc at willowgarage.com>
>
>
> Hey all,
> Ze'ev Klapow spent some time last week implementing a groups feature
> for dynamic reconfigure. The feature enables you to do things like
> show and hide groups of parameters based on what other parameters have
> been set, view parameters for a node by group and expand or collapse
> them in the client, and access parameters in both C++ and Python in a
> consistent manner using '.' syntax. The feature is, I think, super
> useful, but to land in E-Turtle it needs a review. We'd love feedback
> from anyone using dynamic_reconfigure who wants to try out this
> feature.
> The list of changes lives here, along with a section for feedback:
> http://www.ros.org/wiki/dynamic_reconfigure/Reviews/7-1-11_groups
> And updated tutorials live here:
> http://www.ros.org/wiki/Ze%27ev%20Klapow/dynamic
> And the code can be found here:
> https://kforge.ros.org/project/common/services/dynamicreconfig/
> Thanks much, and hope all is well,
> Eitan
> _______________________________________________
> ros-developers mailing list
> ros-developers at code.ros.org
> https://code.ros.org/mailman/listinfo/ros-developers
>