Insights

A General-Purpose Simulator

Combine pieces from previous chapters and clean up a little
A generic Actor as parent for coders, managers, testers, etc.
- Constructor calls a lifecycle method post_init for extra initialization
- Saves the process in case we do want to interrupt it
- Automatically schedules the process to be run
- Provides a .log method for recording events

class Actor(Recorder):
    def __init__(self, sim):
        super().__init__(sim)
        self.post_init()
        proc = self.run()
        self.sim.process(proc)

    def post_init(self):
        pass

    def log(self, state):
        self.sim.log.actor(self.__class__.__name__, self.id, state)

Define a Log class to store collected data

@dataclass
class Log:
    env: Environment | None = None
    queue_events: list = field(default_factory=list)
    actor_events: list = field(default_factory=list)

    def queue(self, name, length):
        self.queue_events.append({"time": self.env.now, "name": name, "length": length})

    def actor(self, kind, id, state):
        self.actor_events.append({"time": self.env.now, "kind": kind, "id": id, "state": state})

Use job fragmentation as discussed in previous chapter
Testers can send jobs back for rework
Or for integration
- In which case every programmer has to do some work

The Parameterization Problem

There are now 15 parameters controlling the behavior of our simulation

class Params:
    t_code_arrival: float = 2.0
    t_code_mean: float = 0.5
    t_code_std: float = 0.6
    t_decomposition: float = 0.5
    t_integration: float = 0.2
    t_interrupt_arrival: float = 5.0
    t_interrupt_mean: float = 0.2
    t_interrupt_std: float = 0.1
    t_queue_monitor: float = 5.0
    t_sim: float = 20
    n_coders: int = 2
    n_iter: int = 1
    n_seed: int = 97531
    n_testers: int = 2
    p_rework: float = 0.5

Some are completely synthetic (number of testers, queue monitoring interval, etc.)
But job and interrupt arrival rates, rework probability, integration times, etc. are not
One of the reasons to build simulations like this is to figure out what numbers we need to know in order to characterize our development process
Limited value in building a more complicated simulation if we can't get numbers for this one