Headless servers

[smubita]

Would it be too much to ask model creators to change their run.py files to take an optional argument using something like sys.argv[], perhaps labelled headless. So that to run the server in a headless mode, one could run

mesa runserver --headlesss project

Then main.py in the mesa project could be modified to look for the --headless option and include it:

    args = [sys.executable, str(run_path), 'headless']
    call(args) 

In turn, run,.py would look for that argument. If the option was present, they could use server.launch() with the open_browser parameter set to false.

server.launch(open_browser = False)

Here is an example: https://github.com/projectmesa/mesa-examples/pull/29/files . If this seems acceptable, an issue could be raised to modify the examples in mesa-examples in a similar fashion and to update documentation in repo mesa accordingly.

This change would allow people to more easily run a model inside a docker container. Or it would be a lighter way of running a long model if you didn't care about seeing the output until the work is done.

[smubita]

If this was done, we could change the Dockerfile (or add a new Dockerfile) that runs it in that same headless mode:

CMD ["sh", "-c", "cd $MODEL_DIR && python3 run.py $HEADLESS"]

The variable $SHEADLESS could default to the empty string.

[rht]

This change would allow people to more easily run a model inside a docker container. Or it would be a lighter way of running a long model if you didn't care about seeing the output until the work is done.

I think it is simpler to have a 1 line change,

server.launch(open_browser=True)

instead of the additional conditional to check the sysarg. Fewer boilerplate lines, and also the argument can be toggled quickly if the model needs to be run on a Docker container.

[smubita]

Changed it accordingly. Thanks for the comments.

[smubita]

Merged in, so I'm closing this.