ROS Resources: Documentation | Support | Discussion Forum | Index | Service Status | ros @ Robotics Stack Exchange
Ask Your Question

Significance of ros::spinOnce()

asked 2011-11-10 03:44:55 -0500

karthik gravatar image

updated 2020-11-21 11:14:57 -0500

lucasw gravatar image

Hi, Can somebody explain what is the actual significance of the ros::spinOnce(). How is it useful? What are the situations where its impact is higher? And also difference between ros::spin() ros::spinOnce().

Thanks, Karthik

edit retag flag offensive close merge delete

2 Answers

Sort by ยป oldest newest most voted

answered 2011-11-10 07:28:34 -0500

Patrick Mihelich gravatar image

The roscpp overview has some background info on callbacks and spinning.

In the background, ROS monitors socket connections for any topics you've subscribed to. When a message arrives, ROS pushes your subscriber callback onto a queue. It does not call it immediately. ROS only processes your callbacks when you tell it to with ros::spinOnce().

This is all part of roscpp's "toolbox, not framework" philosophy. roscpp does not mandate a particular threading model for your node, nor does it demand to wrap your main(). ros::spin() is purely a convenience, a main loop for ROS that repeatedly calls ros::spinOnce() until your node is shut down.

There are a few cases when calling ros::spinOnce() directly is useful.

Processing callbacks at a specific time interval

In rare situations you might want to process ROS callbacks only at certain intervals. Make sure to consider your subscriber queues when doing this. If messages arrive at 100Hz, you call ros::spinOnce() every 5Hz, and your subscriber has a queue_size of 1, you'll drop 95/100 messages! That may or may not be what you want.

Usually a Timer is a better way to inject periodic processing.

Perform processing unsuitable for a ROS callback

Perhaps you are using another library that needs to do its own asynchronous event processing. In that case you need to include its analogue to ros::spinOnce() in your main loop:

ros::Rate r(100);
while (ros::ok())
  libusb_handle_events_timeout(...); // Handle USB events
  ros::spinOnce();                   // Handle ROS events

Integrating ROS with framework X

Many frameworks for GUIs, game programming, etc. do wrap your main(). In that case you just have to play by their rules and find a suitable place to drop in ros::spinOnce().

For example, say you're writing a visualizer using OpenGL and GLUT. One way to integrate ROS would be to tell GLUT to call ros::spinOnce() from its main loop every 10ms:

void timerCb(int value) { ros::spinOnce(); }

glutTimerFunc(10, timerCb, 0);
glutMainLoop(); // Never returns
edit flag offensive delete link more



Is there any use then for a ros::spin() or ros::spinOnce() in the main function of a node that has no callbacks?

2ROS0 gravatar image 2ROS0  ( 2016-07-19 13:22:44 -0500 )edit

I would like to ask the same question also. But I do notice that there ain't no ros node without any topic publishment or subscription...

HaFred gravatar image HaFred  ( 2021-11-26 03:07:50 -0500 )edit

Seems like you only need spin for subscribers, not publishers:

beluga gravatar image beluga  ( 2022-08-04 03:29:11 -0500 )edit

answered 2011-11-10 04:53:51 -0500

DimitriProsser gravatar image

ros::spin() will not return until the node has been shut down. Thus, you have no control over your program while ROS is spinning. The advantage of ros::spinOnce() is that it will check for callbacks/service calls/etc only as often as you call it. Why is this useful? Being able to control the rate at which ROS updates allows you to do certain calculations at a certain frequency.

If, for example, you have an accelerometer that samples at 100Hz. At each update of the accelerometer, it registers a callback in your program. That means that your program is processing the accelerometer data 100 times per second. Let's say you want to read that data and save it to a file five times per second (5Hz). If you're using ros::spin(), your program will only process the callback functions at 100Hz. If you use ros::spinOnce(), you can interject your own code at your desired frequency:

int main()
  ros::Rate loop_rate(5); // 5Hz

  while (ros::ok())

This allows you to execute your own code in between ROS's callback handling.

edit flag offensive delete link more


Thanks DimitriProsser. I partially understand what you explained. Can you tell me what happens if we remove ros::spinOnce() in the above code (having rest the same).
karthik gravatar image karthik  ( 2011-11-10 05:22:50 -0500 )edit
I believe the program will run without checking callbacks/services when spinOnce is removed. When spin is placed instead, the checks will happen at the pace of callback handling. When spinOnce is there then the pace is fixed by user to be 5hz. Let me know if my understanding is correct. Thanks alot.
karthik gravatar image karthik  ( 2011-11-10 05:25:50 -0500 )edit

Without spinOnce() function, you get nothing from ROS (no messages, no services, etc).

Mike Gao gravatar image Mike Gao  ( 2014-11-06 12:13:13 -0500 )edit

@DimitriProsser let's say that at the first 1 second, it samples 100 data : data 1, 2 ,3, ...100. Then They put it on queue. We spin at 5Hz, which means it only calls back 5 times during 1 second. But which data it got at the call backs isn't it data 1,2,3,4,5 since we call from the queue?pls explai

alienmon gravatar image alienmon  ( 2016-08-12 03:11:42 -0500 )edit

The data will not be 1, 2, 3, 4, 5. because the size of the queue is set when you subscribe. For example, when you set the queue to be 10, when the acquired data is already heaped into the queue, the data obtained earlier in the queue will be discarded, which is how the queue works

Dongchan Year gravatar image Dongchan Year  ( 2021-10-13 20:44:19 -0500 )edit

Question Tools



Asked: 2011-11-10 03:44:55 -0500

Seen: 67,171 times

Last updated: Nov 10 '11