Use of agents/ZMQ

[Peque]

@micromouseonline I think you once told me you had a look at my Bulebule's connect_bluetooth.py script and found it confusing with its use of osBrain agents (understandable).

I do not think that should be an issue in the short term and hopefully you can just ignore that part while we advance on this project.

Otherwise, we could consider replacing osBrain/ZMQ use with asyncio.

[micromouseonline]

This would be for network IO? — Peter Harrison peter.harrison@helicron.net

…

On 9 Dec 2019, at 20:43, Miguel Sánchez de León Peque ***@***.***> wrote: @micromouseonline I think you once told me you had a look at my Bulebule's connect_bluetooth.py script and found it confusing with its use of osBrain agents (understandable). I do not think that should be an issue in the short term and hopefully you can just ignore that part while we advance on this project. Otherwise, we could consider replacing osBrain/ZMQ use with asyncio. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or unsubscribe.

[micromouseonline]

I don't think I really get what your code is meant to do.

Why did you choose osbrain to do what appears to be no more than communicate with the robot. What does it bring to the task?

[Peque]

I am just used to osBrain. It allows for easy multiprocessing, easy ZMQ sockets communication and easy remote access to agents.

You can think of the agent as a simple process. In the case of this simple script, the agent is constantly watching for incoming messages from bluetooth/serial and processing the logs. Then you can communicate with the agent from the command line (cmd module). An agent allows you to safely call its methods remotely. So, for example, in order to retrieve some data from the command line, you can just call agent.method() or agent.get_attr("attribute_name"). Behind the scenes there is multithreading to handle these calls safely while the agent keeps doing its other tasks.

Using multiprocessing (or osBrain in this case) is less interesting if you are doing many I/O tasks like we are right now. For those cases multithreading (or Qt threads if we think of a GUI) is probably more useful.

ZMQ is probably not necessary here, since there is only one agent and, therefore, not really lots of communications.

[Peque]

PS: (disclaimer) I created osBrain, so I sometimes just use it because it is simple and straight-forward for me since I already know the tool. That does not mean it is necessarily the best tool for the task. Just the fastest for me to get the job done. 😜