[ros-users] [Discourse.ros.org] [Next Generation ROS] ROS 2 Time vs std::chrono

andreaspasternak ros.discourse at gmail.com
Wed Oct 3 16:06:12 UTC 2018

We were investigating if we should use ROS 2 Time or `std::chrono` for our product. Note that our use case is a highly safety critical real-time system influencing the priority of the arguments.

Upon closer inspection, I found that ROS 2 Time is largely redundant to `std::chrono`.
All functionality of ROS 2 simulated time can also be achieved by providing a custom `std::chrono` clock encapsulating the `rclcpp::Clock` which would look like this (modulo some shared ptr shenanigans I omitted):
struct ros2_clock
  typedef std::chrono::nanoseconds 				duration;
  typedef duration::rep	  				rep;
  typedef duration::period	  				period;
  typedef std::chrono::time_point<ros2_clock, duration> 	time_point;

  static constexpr bool is_steady = false;

  static time_point
  now() noexcept {
    return time_point(duration(clock.now().nanoseconds()));

  static Clock clock;
So from my point of view it is hard to justify using ROS 2 Time as:

1. `rclcpp` uses `std::chrono::Duration` already.
2. `std::chrono` is safer as it is not possible to mix durations from different clocks.
3. `std::chrono` allows a custom time representation allowing to create overflow and underflow safe time operations.
4. With sticking to a single time library there is no confusion on which to use when and no performance degradation from type conversion.
5. `std::chrono` is the C++ standard many people are already familiar with.

So given our use cases, are there arguments speaking for using ROS 2 Time in our code instead of `std::chrono`?

[Visit Topic](https://discourse.ros.org/t/ros-2-time-vs-std-chrono/6293/1) or reply to this email to respond.

More information about the ros-users mailing list