ros::Time::isValid doesn't do what we think it does?

Question

From the documentation (http://docs.ros.org/melodic/api/rostime/html/classros_1_1Time.html#a84cd12a499c70fb3eefce89cda2b6113), ros::Time::isValid returns "whether or not the current time is valid. Time is valid if it is non-zero".

However, a trivial example is enough to show that the implementation does not do this:

#include <ros/ros.h>

int main(int argc, char** argv) {
    ros::init(argc, argv, "rostime_example");
    ros::NodeHandle n;

    // default construct a time
    ros::Time t;

    ROS_WARN_STREAM("time value: " <<t.toSec());
    ROS_WARN_STREAM("time valid: " <<t.isValid());

    return 0;
}

The output is:

ivaughn@ros-pc:~/test_ws$ rosrun rostime_example node 
[ WARN] [1561399516.961348750]: time value: 0
[ WARN] [1561399516.961420491]: time valid: 1

Any idea what's going on here? Is this just a bug in ROS 1, or am I misunderstanding how this function is expected to behave.

We're kind of curious, behavior that does something different when a timestamp is invalid is often safety-critical. Thanks!

---

Originally posted by ivaughn_whoi on ROS Answers with karma: 13 on 2019-06-24

Post score: 1

Answer 1

I believe this is due to a misunderstanding of the comment on isValid().

The documentation indeed states (here):

Returns whether or not the current time is valid. Time is valid if it is non-zero.

But "the [..] time" it refers to is not the Time instance, but "the time", as in: the global ROS time.

A ROS node can use different methods to retrieve the "current time": it can use walltime, a simulation time or something else. The "valid" here essentially returns true whenever one of those sources have been selected and has been initialised properly.

0 is chosen as a sentinel value to detect when that is not the case (as it's assumed that 0 is just not a valid value for any source of time (whether that is correct is something else)).

It's slightly confusing that a ros::Time instance can also have a 0 value and that ros::Time::isValid() is a static method of the ros::Time class.

In the code that you show:

int main(int argc, char** argv) {

    ros::init(argc, argv, "rostime_example");
    [..]
}

isValid() will return true as soon as you call ros::init(..) (as one of its duties is to initialise the time subsystem).

---

We can also take a look at the sources (from here):

bool Time::isValid()
{
  return (!g_use_sim_time) || !g_sim_time.isZero();
}

This essentially tells us that "time" is considered "valid" if either we're not using simulation time, or, if simulation time is being used, it's != 0.

The g_ prefixes tell us that this static method is checking globals, not ros::Time member variables.

---

Edit: perhaps a nice contribution would be to improve the documentation of this method and to remove the ambiguity.

---

Originally posted by gvdhoorn with karma: 86574 on 2019-06-24

This answer was ACCEPTED on the original site

Post score: 2

---

Original comments

Comment by gvdhoorn on 2019-06-24:
And this is not directed at you, but:

Is this just a bug in ROS 1, [..]

I like how posters here often insert these sort of statements almost casually in their questions.

Comment by ivaughn_whoi on 2019-06-24:
Yeah, I'd like to make sure I correctly understand what's going on before claiming its a bug. WIth ROS its more likely to be a documentation issue than anything else.

Pull request: https://github.com/ros/roscpp_core/pull/109