Deterministic Code Generation

Most code generation tools focus on producing code.

This project is about something else:

preserving architecture.

My book "Raising the Level of Abstraction" provides all the details.

---

The Problem

Many systems start clean and understandable, then degrade over time.

Not because the code is bad, but because the structure is not enforced.

• components grow organically
• interfaces drift
• patterns are copied inconsistently
• architectural intent is lost

The result is a system that works, but cannot be reasoned about.

---

The Idea

Instead of writing system structure directly in C++, we describe it at a higher level:

• what data exists
• how that data flows
• which components process it

Then we generate the implementation.

The generator is deterministic:

• same input → same output
• no hidden behavior
• no runtime framework

The generated code is just plain C++.

---

What This Repository Contains

This repository contains two example generators:

SimpleGenerator

A minimal example showing:

• YAML model input
• Jinja2 templates
• generated C++ types

This is the starting point.

---

SerializationGenerator

Builds on the simple model to generate:

• measurement types
• binary serialization (BDS-style)
• debugging/logging (operator<<)

This demonstrates how repetitive, error-prone code can be generated consistently.

---

Why This Approach

This is not an IDL.

It does not generate code for a runtime system.

Instead:

• the model describes the system itself
• the generator produces final implementation code
• the result is explicit, readable, and debuggable

There is:

• no reflection
• no schema interpreter
• no hidden allocation

Just C++.

---

Example

A small model:

measurements:
  - Temperature
  - Humidity

queues:
  - TemperatureQueue

stages:
  - TemperatureSensor
  - Logger

connections:
  - from: TemperatureSensor
    to: TemperatureQueue
  - from: TemperatureQueue
    to: Logger

## Getting Started
1. Install dependencies
      pip install pyyaml jinja2
2. Run a generator
      cd SimpleGenerator
      ./generator.py

Generated code will be written to the output directory.