NetLogo-like concise API for Mesa model specification

[rht]

The way default logging object works, and the way get_default_rng in #2021 2021, reminds me of the API of Pennylane, a Python library for quantum machine learning.

If you define this decorator

# Decorator to inject 'self' into functions
def inject_self(func):
    @wraps(func)
    def wrapper(self, *args, **kwargs):
        # Inspect the local variables of the function
        local_vars = inspect.currentframe().f_back.f_locals

        # Find all callables in the scope of the function
        functions = {k: v for k, v in local_vars.items() if callable(v)}

        # Wrap each function so it accepts 'self'
        for name, function in functions.items():
            local_vars[name] = partial(function, self)

        # Call the original function
        return func(self, *args, **kwargs)
    return wrapper

It allows you to do this in a model step (note: I just tried, I didn't even have to modify the source code of the method do in AgentSet. I just need to move it up to be a global function)

@inject_self
def step(self):
    shuffle().do("move")
    do("trade")
    _advance_time()
    do("sample_random")

This is similar to Solara @solara.component injector works, in that it injects parent context into the child subcomponents.

You could possibly go further, and specify your model in a way similar to AgentPy, but briefer (cc: @jofmi)

@mesa.step
def step(self):
    shuffle().do("move")
    _advance_time()

params = {
  "num_agents": 10,
  "fraction_minority": 0.3
}

# The datacollector is specified elsewhere
@mesa.model
def model(params, datacollector):
    initialize_agents()
    collect()
    for i in range(100):
        step()
        collect()

If there is a consensus that this is worth experimenting, I could build it and have it in mesa.experimental.

For reference, here is how a quantum circuit is specified in Pennylane:

@qml.qnode(dev, diff_method="finite-diff")
def circuit(params):
    # define your quantum circuit here
    qml.RX(params[0], wires=0)
    qml.CNOT(wires=(0, 1))
    qml.RY(params[1], wires=1)
    qml.RZ(params[2], wires=1)
    qml.RX(params[3], wires=1),
    return qml.state()

where each of the function calls is injected with the @qml.qnode's arguments.