Message named pkg/Duration with field builtin_interfaces/Duration is invalid

[peci1]

Generated by Generative AI

No response

Operating System:

Linux

ROS version or commit hash:

rolling

RMW implementation (if applicable):

No response

RMW Configuration (if applicable):

No response

Client library (if applicable):

No response

'ros2 doctor --report' output

No response

Steps to reproduce issue

1. Create a message package pkg containing a message type named Duration with a field of type builtin_interfaces/Duration.
2. Compile the workspace and inspect the generated Python code.
3. Run mypy tests and flake8

Expected behavior

Message type pkg/Duration should be valid.

Actual behavior

The generated message Python code is invalid:

 27: /root/target_ws/build/stamped_msgs/rosidl_generator_py/stamped_msgs/msg/_duration.py:87:1: F811 redefinition of unused 'Duration' from line 20
  27: class Duration(rosidl_pycommon.interface_base_classes.BaseMessage, metaclass=Metaclass_Duration):
  27: ^
  27: 
  27: 1     F811 redefinition of unused 'Duration' from line 20

  30: /root/target_ws/build/stamped_msgs/rosidl_generator_py/stamped_msgs/msg/_duration.py:87:1: error: Name "Duration" already defined (possibly by an import)  [no-redef]
  30: /root/target_ws/build/stamped_msgs/rosidl_generator_py/stamped_msgs/msg/_duration.py: note: In member "__eq__" of class "Duration":
  30: /root/target_ws/build/stamped_msgs/rosidl_generator_py/stamped_msgs/msg/_duration.py:151:27: error: "Duration" has no attribute "header"  [attr-defined]
  30: /root/target_ws/build/stamped_msgs/rosidl_generator_py/stamped_msgs/msg/_duration.py:153:25: error: "Duration" has no attribute "data"  [attr-defined]
  30: Found 3 errors in 1 file (checked 46 source files)

Additional information

Indeed, there are multiple places in the generated code that do:

from a.msg import b

But if the message is named pkg.Duration and the type of one of its fields is builtin_interfaces.Duration, then this import shadows the actual pkg.Duration and the code gets either confused, or even invalid.

I think all usages of from a.msg import b should be replaced by import a.msg; .... a.msg.b in the generated code. This way, the name collisions would not happen.

[peci1]

I've marked a few places which introduce this problem in #206 (review) which is the greates contributor of the new naming constraints (however, some of the restrictions have already been there before this PR).

[InvincibleRMC]

I agree we just need to switch to absolute imports everywhere.

[peci1]

Thanks for confirming. Does anyone have cycles to get this done before Lyrical freeze? Unfortunately, I won't have time until the last week of April, which is probably too late...

[InvincibleRMC]

I don't think I have time to change everything over to absolute, but I have a pr in progress which should resolve the bug from #206. Which mostly had to do with global namespaces. The local namespace behavior was largely unchanged in #206

[InvincibleRMC]

@peci1 Do you have a chance to test if #258 works for your use case? It should resolve global namespace pollution. Most of the local stuff existed before. Testing it locally it worked with a copy of your Duration failure.

[peci1]

Thanks, #258 fixed the most important problems I had. There are still some local pollutants that would also be nice to be absolutized...

[christophebedard]

We still have until April 20th until the freeze (since it was delayed), so we still have time to fix what's left

[peci1]

Sure! However, I'm leaving for vacation in a few hours, so I won't be able to help...