lifecycle_publisher: method naming

[wieset]

Probably not the most important issue but I felt like inquiring about it :)

Is there a rationale behind naming the methods on_activate() and on_deactivate() with the "on" prefix in LifecyclePublisherInterface?

To me, it feels wrong to call these methods directly. They feel like state transition methods that should be called internally when a state transition is requested. Just like in LifecycleNodeInterface I would expect to have methods activate() and deactivate() which would then trigger the state transition.

This is probably loosely related to #654. A universal lifecycle interface for nodes and helper classes with automatic triggering of state transitions would be great!

[Karsten1987]

If I understand you correctly, you'd wish for rclcpp_lifecycle to call these functions implicitly so that the user doesn't have to call this on its own?
e.g. https://github.com/ros2/demos/blob/master/lifecycle/src/lifecycle_talker.cpp#L146-L152

That might not be a trivial change as we have to introduce some book keeping in the lifecycle node to know which lifecycle publishers exist. Also I am wondering if that behavior is wished for in 100% of the cases. But I guess if not, users can always fall back to a regular publisher in that case.

---

Alternatively, we could also rename the methods on the lifecycle publisher interface from on_activate to activate to make it clear that this functions are to be called explicitly.

[wieset]

Thanks for your answer! Your first point is the one raised in #654 and would of course be nice to have, but would need some development effort. For now, I was mainly thinking of your second point, renaming the methods to activate() / deactivate(). I think that would make their intent clearer - to be called by the user to activate / deactivate the publisher.