launch testing : Wait for topics to publish

[adityapande-1995]

This aims to close #272. It aims to provide an API in to be used with launch_testing where we provide a set of topics and we wait for at least a single message to appear on all of them.

Sample usage :

import unittest
import launch
import launch.actions
import launch_ros.actions
import launch_testing.actions
import launch_testing.markers
import pytest
import rclpy
from rclpy.node import Node
from std_msgs.msg import String

from launch_testing_ros import WaitForTopics

def generate_node(i):
    return launch_ros.actions.Node(
        executable='talker',
        package='demo_nodes_cpp',
        name='demo_node_' + str(i),
        remappings = [('chatter', 'chatter_' + str(i))]
    )

@pytest.mark.launch_test
@launch_testing.markers.keep_alive
def generate_test_description():
    # 'n' changes the number of nodes and topics
    n = 5
    description = [generate_node(i) for i in range(n)] + [launch_testing.actions.ReadyToTest()]
    return launch.LaunchDescription(description), {'count':n}

class TestFixture(unittest.TestCase):

    def test_check_if_msgs_published(self, proc_output, count):

        topic_list = [('chatter_' + str(i), String) for i in range(count)]
        with WaitForTopics(topic_list, timeout=5.0):
            print("The topics are alive !")

[jacobperron]

I think this is a good start. I've left some feedback below.

Also, I'm not sure that creating a separate node class adds much value; I would have just put all the logic inside WaitForTopics. I'm okay with it either way though.

[jacobperron]

Alternatively, we can use functools.partial.

functools.partial(self.topic_callback, topic_name)

and then the callback would look like:

def topic_callback(self, topic_name, data):
    ...

I don't think there's anything wrong with this, so feel free to leave it as-is.

[jacobperron]

LGTM, two more comments below

[adityapande-1995]

CI :
build : --packages-above-and-dependencies launch_testing_ros
test : --packages-above launch_testing_ros

• Linux
• Linux-aarch64
• macOS
• Windows

[jacobperron]

A couple suggestions for the test code. LGTM

[Blast545]

👨‍🌾 It seems that this PR caused several regressions in the WIndows buildfarms jobs:

nightly_win_rel#2079
nightly_win_deb#2134

Do you think you can take a look to a possible fix @adityapande-1995? I think it's likely related to these lines:

https://github.com/ros2/launch_ros/blob/master/launch_testing_ros/launch_testing_ros/wait_for_topics.py#L20-L23

It seems to me that rclpy can't be imported there for some reason. I'll open a revert PR in the meantime to be sure this PR caused the regressions.

[clalancette]

It's weird that this caused a regression, particularly because CI seems to have been run properly and came back green. But let's see if the revert PR fixes it.

[Blast545]

I can't replicate the error in CI. I'll investigate a bit deeper before reaching any conclusion.

[aprotyas]

It seems to me that rclpy can't be imported there for some reason. I'll open a revert PR in the meantime to be sure this PR caused the regressions.

I believe that's for this. Not sure why it can't be imported before running tests though.
https://github.com/ros2/rclpy/blob/6dd9540661631c971499d5672fb50a8ec1e3ac22/rclpy/test/__init__.py#L20