[ros-users] ros::message_traits::TimeStamp without using Header

[Sebastian Schneider]

Hello,

I recently started using the message_filters::Cache and it seems that 
only messages, containing a Header element, can be used out of the box.

For what I found out, if I place a Header element in the message 
definition file (Something.msg) the generated C++ header file 
automatically contains the required function:

static ros::Time ros::message_traits::TimeStamp<M>::value( const M& m );

If I don't include the Header element, this function does not exist and 
the Cache cannot be used.

For my personal needs I don't require all the information in the Header 
element. A simple time element would suffice. Also, I don't see why the 
cache would need more than a simple timestamp. But, from what I found 
out, if I replace the Header element with a simple time element, the 
required message traits are not generated, thus the Cache doesn't work 
(or even compile).

As a workaround I found out, that I can define the required message 
traits myself like so:

Assuming my message definition file (Something.msg) starts with:

time stamp
...

If I put the following code fragment somewhere in my code before I start 
using the Cache, it's working:

namespace ros
{
     namespace message_traits
     {
         template<>
         struct TimeStamp<Something>
         {
             static ros::Time* pointer( Something& m )
             {
                 return &m.stamp;
             }

             static ros::Time const* pointer( const Something& m )
             {
                 return &m.stamp;
             }

             static ros::Time value( const Something& m )
             {
                 return m.stamp;
             }
         };
     } // namespace message_traits
} // namespace ros

Finally, my question: Is there a way to let the shown code fragment be 
generated automatically as it will be if I used the Header element? Or 
to put it differently: Can I somehow mark a time element in the message 
definition file, such that the code fragment is automatically appended 
to the generated c++ header?

Thanks,

Sebastian