fix(set_message): encode strings if field is bytes

[russkel]

I was trying to load yaml that was output from ros2 topic echo and then convert them back into msgs for fixtures, and I came across this issue:

Traceback (most recent call last):
  File "/opt/ros/jazzy/lib/python3.12/site-packages/rosidl_runtime_py/set_message.py", line 85, in set_message_fields_internal
    value = field_type(field_value)
            ^^^^^^^^^^^^^^^^^^^^^^^
TypeError: string argument without an encoding

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "/opt/ros/jazzy/lib/python3.12/site-packages/rosidl_runtime_py/set_message.py", line 58, in set_message_fields_internal
    items = values.items()
            ^^^^^^^^^^^^
AttributeError: 'str' object has no attribute 'items'

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "/opt/ros/jazzy/lib/python3.12/site-packages/rosidl_runtime_py/set_message.py", line 101, in set_message_fields
    set_message_fields_internal(msg, values, timestamp_fields)
  File "/opt/ros/jazzy/lib/python3.12/site-packages/rosidl_runtime_py/set_message.py", line 97, in set_message_fields_internal
    set_message_fields_internal(
  File "/opt/ros/jazzy/lib/python3.12/site-packages/rosidl_runtime_py/set_message.py", line 88, in set_message_fields_internal
    set_message_fields_internal(
  File "/opt/ros/jazzy/lib/python3.12/site-packages/rosidl_runtime_py/set_message.py", line 60, in set_message_fields_internal
    raise TypeError(
TypeError: Value '' is expected to be a dictionary but is a str

With the following yaml:

header:
  stamp:
    sec: 1730513229
    nanosec: 6167372
  frame_id: ''
status:
- level: "\x02"
  name: 'geofence_node: Go GeoFences'
  message: No Geofences
  hardware_id: none
  values: []
- level: "\0"
  name: 'geofence_node: No Go GeoFences'
  message: Outside All
  hardware_id: none
  values: []

The level value is a string \x02 but set_message is trying to convert it to the field_type bytes. This patch fixes this issue by encoding the string to bytes.

[russkel]

If this could be backported to jazzy that would be lovely too.

[fujitatomoya]

looks good to me but i would like to have 2nd review just in case something missing.

[fujitatomoya]

Pulls: #30
Gist: https://gist.githubusercontent.com/fujitatomoya/451def5286e30198246798264ec9e99a/raw/5973684a8f6b8eb319279f18f63da329c50b5e5e/ros2.repos
BUILD args: --packages-above-and-dependencies rosidl_runtime_py
TEST args: --packages-above rosidl_runtime_py
ROS Distro: rolling
Job: ci_launcher
ci_launcher ran: https://ci.ros2.org/job/ci_launcher/14779

• Linux
• Linux-aarch64
• Linux-rhel
• Windows

[russkel]

I added a basic smoke test for conversion from yaml using one of the test message fixtures and it appears not good enough with a string containing \xff. Although maybe this is an issue with the way some bytes are encoded into yaml/strings and become irrecoverable...

In [1]: moo = '\xff'

In [2]: moo
Out[2]: 'ÿ'

In [3]: moo.encode()
Out[3]: b'\xc3\xbf'

In [9]: moo = chr(0xff)

In [10]: moo
Out[10]: 'ÿ'

In [11]: moo.encode(encoding='utf8')
Out[11]: b'\xc3\xbf'

In [12]: moo.encode(encoding='utf16')
Out[12]: b'\xff\xfe\xff\x00'

In [13]: moo.encode(encoding='ascii')
---------------------------------------------------------------------------
UnicodeEncodeError                        Traceback (most recent call last)
Cell In[13], line 1
----> 1 moo.encode(encoding='ascii')

UnicodeEncodeError: 'ascii' codec can't encode character '\xff' in position 0: ordinal not in range(128)

In [17]: for c in moo: print(ord(c))
255

In [19]: import yaml

In [20]: yaml.dump('\xff')
Out[20]: '"\\xFF"\n'

In [21]: from rosidl_runtime_py.convert import _convert_value

In [22]: _convert_value('\xff')
Out[22]: 'ÿ'

In [27]: bytes(moo, 'ascii', 'backslashreplace')
Out[27]: b'\\xff'

[russkel]

The only way I am able to get this to pass is using value = bytes([ord(c) for c in field_value]).
I suspect this isn't terribly correct.... 🤔

[cottsay]

Does passing canonical=True to yaml.dump() do what you need (possibly here)?