ROS Resources: Documentation | Support | Discussion Forum | Index | Service Status | Q&A
Ask Your Question

¿How does the queue size and callback mechanism exactly work?

asked 2022-03-21 07:25:03 -0600

IgnacioUD gravatar image

updated 2022-04-17 10:53:34 -0600

lucasw gravatar image

[Working in ROS noetic ubuntu 20.04] I've a topic publishing robot joints at 100 Hz and the following code for testing purposes:

from sensor_msgs.msg import JointState
import rospy

rospy.init_node("queue_size_testing", anonymous=True)

def callback(msg) -> None:
    now =
    print(f"[ACTUAL] secs={now.secs} nsecs={now.nsecs} [READ]secs={msg.header.stamp.secs} nsecs={msg.header.stamp.nsecs}")


I'm just subscribing to the topic and executing the callback at 10 Hz (obviously this code is not appropriate since it makes no sense to make the callback sleep, it is developed for testing purposes). So, this code is able to process 10 messages per second, even if 100 are coming from the publisher. As I have not defined a value for the subscriber's queue_size, it has taken its default value (infinite). I̶n̶ ̶t̶h̶i̶s̶ ̶w̶a̶y̶,̶ ̶I̶ ̶u̶n̶d̶e̶r̶s̶t̶a̶n̶d̶ ̶t̶h̶a̶t̶ ̶t̶h̶e̶ ̶s̶u̶b̶s̶c̶r̶i̶b̶e̶r̶ ̶q̶u̶e̶u̶e̶ ̶s̶h̶o̶u̶l̶d̶ ̶g̶r̶o̶w̶ ̶a̶t̶ ̶a̶ ̶r̶a̶t̶e̶ ̶o̶f̶ ̶9̶0̶ ̶m̶e̶s̶s̶a̶g̶e̶s̶ ̶p̶e̶r̶ ̶s̶e̶c̶o̶n̶d̶ ̶a̶n̶d̶ ̶s̶h̶o̶u̶l̶d̶ ̶d̶i̶s̶p̶l̶a̶y̶ ̶o̶n̶ ̶t̶h̶e̶ ̶s̶c̶r̶e̶e̶n̶ ̶a̶l̶l̶ ̶t̶h̶e̶ ̶m̶e̶s̶s̶a̶g̶e̶s̶ ̶r̶e̶c̶e̶i̶v̶e̶d̶ ̶(̶w̶i̶t̶h̶o̶u̶t̶ ̶a̶n̶y̶ ̶m̶s̶g̶s̶ ̶d̶r̶o̶p̶p̶i̶n̶g̶)̶,̶ ̶s̶i̶n̶c̶e̶ ̶i̶t̶ ̶r̶e̶s̶p̶e̶c̶t̶s̶ ̶t̶h̶e̶ ̶F̶I̶F̶O̶ ̶p̶h̶i̶l̶o̶s̶o̶p̶h̶y̶. (This doesn't happen as if the queue_size is 10 and there are 20 available in the socket stream, the oldest 10 are discarded, the queue_size name for the subscriber may be a little misleading for this reason.) Here is the actual output:

[ACTUAL] secs=308 nsecs=8000000 [READ]secs=297 nsecs=835000000
[ACTUAL] secs=308 nsecs=106000000 [READ]secs=308 nsecs=100000000
[ACTUAL] secs=308 nsecs=108000000 [READ]secs=297 nsecs=845000000
[ACTUAL] secs=308 nsecs=206000000 [READ]secs=308 nsecs=200000000
[ACTUAL] secs=308 nsecs=208000000 [READ]secs=297 nsecs=855000000
[ACTUAL] secs=308 nsecs=306000000 [READ]secs=308 nsecs=300000000
[ACTUAL] secs=308 nsecs=308000000 [READ]secs=297 nsecs=865000000
[ACTUAL] secs=308 nsecs=406000000 [READ]secs=308 nsecs=400000000
[ACTUAL] secs=308 nsecs=408000000 [READ]secs=297 nsecs=875000000
[ACTUAL] secs=308 nsecs=506000000 [READ]secs=308 nsecs=500000000
[ACTUAL] secs=308 nsecs=508000000 [READ]secs=297 nsecs=885000000
[ACTUAL] secs=308 nsecs=606000000 [READ]secs=308 nsecs=600000000
[ACTUAL] secs=308 nsecs ...
edit retag flag offensive close merge delete

1 Answer

Sort by » oldest newest most voted

answered 2022-03-25 11:08:12 -0600

miura gravatar image

Are there multiple subscribers running in parallel?

I would expect that being able to lock would eliminate multiple callbacks running at the same time.

Or if we could output the thread identifier, we would be able to see the movement.

edit flag offensive delete link more


A single publisher and subscriber were running at a time, there was no further code only the one I shared, that's why I find it strange.

IgnacioUD gravatar image IgnacioUD  ( 2022-03-28 03:16:58 -0600 )edit

I see. I don't have any other ideas at the moment. I hope someone can give me a good answer.

miura gravatar image miura  ( 2022-03-28 08:00:22 -0600 )edit

Your Answer

Please start posting anonymously - your entry will be published after you log in or create a new account.

Add Answer

Question Tools



Asked: 2022-03-21 07:25:03 -0600

Seen: 138 times

Last updated: Mar 25 '22