ros::this_node::getNamespace() ought to return sanitized namespace

[lucasw]

  ros::init(argc, argv, "ns_test");
  std::cout << ros::this_node::getNamespace() << std::endl;
  ros::NodeHandle nh("~");
  std::cout << nh.getNamespace() << std::endl;

This is fine

$ rosrun ns_test ns_test __ns:=foo
/foo
/foo/ns_test

This ought to print the same

rosrun ns_test ns_test __ns:=/foo
//foo
/foo/ns_test

An additional / was prefixed to the provided namespace.

and this (which is somewhat ridiculous, but why not fix it also)

rosrun ns_test ns_test __ns:=//foo//
///foo//
/foo/ns_test

The namespace gets cleaned up getting into the NodeHandle, but the this_node returns exactly what was on the command line plus the additional forward slash.

Using the released kinetic version currently.

[wjwwood]

Makes sense to me, though it might be someone is relying on this behavior (they want what came in over the CLI, not the sanitized version), but we can definitely change it for newer distributions and consider doing it anyways for existing distributions.

It's definitely worth making a pr against lunar-devel if you have the time.

[lucasw]

I notice roslaunch

<node name="ns_test"
    pkg="ns_test"
    type="ns_test"
    output="screen"
    ns="//foo//"/>

prints out

NODES
  //foo//
    ns_test (ns_test/ns_test)
...
process[/foo//ns_test-1]: started with pid [26179]
...
[/foo//ns_test-1] process has finished cleanly

The NODES part prints the namespace verbatim, while the process messages are removing a leading forward slash.

[lucasw]

If there were to be a namespace cleaning test test/test_roscpp/test/launch/namespace.xml and test/test_roscpp/test/src/namespaces.cpp could be the basis for it.

(and build and test with

catkin_make test_roscpp-namespaces-clean
rostest src/ros_comm/test/test_roscpp/test/launch/namespaces-clean.xml

)

[dirk-thomas]

The argument __ns is not sanitized at the moment and the user is expected to pass a non-global namespace without a trailing slash.

I am not sure why you want to pass namespaces with leading and trailing namespaces in the first place but since the provided PR is straight forward it shouldn't be a problem to do the extra sanitization you proposed. Thanks for create the patch.

[dirk-thomas]

Addressed by #1100.

[k-okada]

I ran into similar situation and still confused that on python , we'll get

$ ROS_NAMESPACE=/hoge/ python -c 'import rospy; print(rospy.get_namespace())'
/hoge/
$ ROS_NAMESPACE=/hoge python -c 'import rospy; print(rospy.get_namespace())'
/hoge/
$ ROS_NAMESPACE=hoge python -c 'import rospy; print(rospy.get_namespace())'
/hoge/
$ ROS_NAMESPACE=hoge/ python -c 'import rospy; print(rospy.get_namespace())'
/hoge/

on rospy, but on roscpp (after applied this patch), we'll get

$ ROS_NAMESPACE=/hoge/ ../../devel/lib/test_namespace/test_namespace
[ INFO] [1501025799.775532508]: namespace -> /hoge
$ ROS_NAMESPACE=/hoge ../../devel/lib/test_namespace/test_namespace
[ INFO] [1501025804.109156084]: namespace -> /hoge
$ ROS_NAMESPACE=hoge ../../devel/lib/test_namespace/test_namespace
[ INFO] [1501025806.913568473]: namespace -> /hoge
$ ROS_NAMESPACE=hoge/ ../../devel/lib/test_namespace/test_namespace
[ INFO] [1501025810.195600461]: namespace -> /hoge

the source code is

$ cat src/test_namespace.cpp
#include <ros/ros.h>


int main(int argc, char **argv) {
  ros::init(argc, argv, "test_namespace");
  ROS_INFO_STREAM("namespace -> " <<  ros::this_node::getNamespace());
}

so question is, do we have tailing '/' on namespace or not.