Hi,
first of all, thanks for this super-useful package!

I would like to highlight some compatibility issue with Python3 and how I managed to solve them.

Import statement

The import statement shown in the README causes an AttributeError, specifically:

AttributeError: module 'ros_np_multiarray' has no attribute 'to_multiarray_f32'

Instead, one should use

from ros_np_multiarray import ros_np_multiarray as rnm

and everything works fine.

Conversion to NumPy

When running one of the to_numpy, a TypeError occurs, specifically:

TypeError: expected sequence object with len >= 0 or a single integer

The issue can be solved by modifying the first line of the _multiarray_to_numpy function to convert dims to a tuple, i.e.

dims = tuple(map(lambda x: x.size, multiarray.layout.dim))

As the mentioned changes seem to work in Python2.7 as well, it could be useful to update the code in order to achieve Python3 compatibility.

System's info:

• Ubuntu 18.04-LTS
• ROS Melodic

Hi,
I am trying to use this package to send np arrays using ROS2 but got this error:

[inference.py-1]   File "/home/docker_gnn/colcon_ws/install/_gnn/lib/_gnn/inference.py", line 363, in lidar_callback
[inference.py-1]     cur_msg.array = rnm.to_multiarray_f32(cur_traj.astype('float32'))
[inference.py-1]   File "/home/docker_gnn/.local/lib/python3.10/site-packages/ros_np_multiarray/ros_np_multiarray.py", line 11, in _numpy_to_multiarray
[inference.py-1]     multiarray.layout.dim = [MultiArrayDimension('dim%d' % i,
[inference.py-1]   File "/home/docker_gnn/.local/lib/python3.10/site-packages/ros_np_multiarray/ros_np_multiarray.py", line 11, in <listcomp>
[inference.py-1]     multiarray.layout.dim = [MultiArrayDimension('dim%d' % i,
[inference.py-1] TypeError: MultiArrayDimension.__init__() takes 1 positional argument but 4 were given

PS: Everything works fine with ROS1
Any suggestions?
Thanks