Using the XMILE Compiler Standalone
system dynamics, systemdynamics, xmile, bptk, bptk-py, python, business simulation, stella
Using the XMILE Compiler Standalone
A scenario file with a source compiles its XMILE model for you, as XMILE Step by Step shows. The compiler can also be called on its own - to look at the code it writes, to keep the compiled model in your project, or to run a model many times without scenarios around it.
The compiler writes a file, so the code on this page is shown rather than run. The model is the customer acquisition model of the XMILE quickstart.
Compiling a model
compile_xmile takes the XMILE file, the file to write and a target:
from BPTK_Py.sdcompiler.compile import compile_xmile
compile_xmile("simulation_models/customer_acquisition_xmile.stmx",
"simulation_models/customer_acquisition_xmile.py",
"py-sddsl")There are two targets:
| Target | Writes |
|---|---|
py-sddsl |
A Python file with the model as SD DSL code: a subclass of Model |
json-rust-engine |
The JSON the Rust engine loads - the same model, serialized |
Whatever the SD DSL cannot express is refused with a message naming the element; the XMILE in the SD DSL chapter lists what that is.
What the compiler writes
The py-sddsl file is SD DSL code as you would write it by hand: one element per stock, flow, converter and constant, and their equations. Its beginning:
from BPTK_Py import Model
from BPTK_Py import sd_functions as sd
class CustomerAcquisitionXmile(Model):
def __init__(self):
super().__init__(starttime=0.0, stoptime=60.0, dt=1.0, name='customer_acquisition_xmile')
customers = self.stock('customers')
potentialCustomers = self.stock('potentialCustomers')
customerAcquisition = self.flow('customerAcquisition')
...
customerAcquisition.equation = sd.max(0.0, (acquisitionThroughAdvertising + acquisitionThroughWordOfMouth))
...The names are those of the XMILE model in camel case.
Running it
The class is an ordinary Model, so simulate runs it and returns a DataFrame - on the Python engine by default, on the Rust engine with backend="rust":
from simulation_models.customer_acquisition_xmile import CustomerAcquisitionXmile
model = CustomerAcquisitionXmile()
model.simulate(["customers", "customerAcquisition"], backend="rust")| t | customers | customerAcquisition |
|---|---|---|
| 0 | 0.0 | 1000.0 |
| 10 | 13843.5 | 1834.2 |
| 20 | 33620.5 | 1917.8 |
| 30 | 49222.2 | 1063.8 |
| 60 | 59686.5 | 36.4 |
Both engines give the same numbers. Everything else the SD DSL offers works on the compiled model as well - changing a constant, plotting an element, registering it with bptk - as the SD DSL chapter describes.