Killing the subprocess after using the visualizer

[nikhilpodila]

I notice two issues with using meshcat to run my visualizations for simulations developed on pinocchio:

1. I cannot effectively kill a meshcat visualizer completely, and thus, its server. Vizualizer().delete() just deletes the path and Visualizer().close() fails since ViewerWindow().close() has not been implemented. This would not be a problem if issue 2 below is addressed.

2. I use Meshcat visualizer by calling Visualizer(zmq_url=None, window=None). When I do so, the ViewerWindow() internally creates a zmq_url, zmq_socket and a subprocess. Once I am done using the visualizer, I would like to delete it, end the program, start a new program and create a new visualizer. But by doing this, I am assigned a new URL and subprocess every time, resulting in a pile up in my system's processes. I see that this is because the subprocess is not killed when I end the program or delete the visualizer. It would be ideal if killing the subprocess and closing the socket could be handled by Meshcat or by calling a Meshcat method such as Visualizer().close()

[kingjin94]

Second that. Thanks to viewer.to_static_html() one can quite comfortably save debug visualizations in bigger optimization runs but as there is no way to free the meshcat visualizer one can also create hughe memory leaks...

[kingjin94]

IMHO I think that the atexit callback in https://github.com/rdeits/meshcat-python/blob/763ac31c869156921fd9560b2bce3761dbbc2cfa/src/meshcat/servers/zmqserver.py#L74 is not the python way to go. Rather would expect resources to be cleaned with gc, i.e. the cleanup routine should be called in __del__ (or an explicit close function called by Visualizer.close() (or ViewerWindow.close()

[kingjin94]

Also zmq does not seem to be handled gracefully according to https://stackoverflow.com/questions/26224179/zmq-socket-gracefully-termination-in-python

Suggestion seems to be to use finally clause to run context.destroy() / term() at the end of main