jmjrawlings / unconstrained

Python devcontainer for constraint programming

Unconstrained

My personal development environment for modelling and solving constraint programming problems in python.

Features

• A fully featured devcontainer
  • Python
  • MiniZinc
  • Google ORTools
• A test dockerfile target (no dev dependencies)
• A prod dockerfile target (no test dependencies)
• Some example models (need more)
  • N-Queens
  • Nurse Rostering
• A strongly typed wrapper for minizinc model results
• A strongly typed Map class (dict replacement)
• A strongly typed List class (list replacement)
• Convenience functions for defining attrs models
• CI/CD scripts using dagger-io
  • Deployed to Github
• Multi layered python package dependency management using pip-tools

Repository Structure

.
├── examples       # Example problems   
├── scripts        # Helper scripts
├── requirements   # Python requirement files
├── build          # Build scripts
├── unconstrained  # Source code
├── tests          # Test suite
├── Dockerfile             
├── pytest.ini              
├── LICENSE.md              
└── README.md               

Links

Constraint Programming

• MiniZinc
• MiniZinc Manual
• MiniZinc Reference
• MiniZinc Basic Modelling Course
• MiniZinc Advanced Modelling Course
• Discrete Algorithms Course
• Google OR-Tools
• Google OR-Tools GitHub
• Google OR-Tools Manual

Python Packages

• Altair
• Attrs
• Cattrs
• OR-Tools
• Minizinc
• Pytest

Development Tools

• Visual Studio Code
• VSCode Devcontainer
• Dagger

Notes / TODO

Docker

• apt-get and apt-install takes forever which is very annoying when playing around with Dockerfiles. Presumably theres a way to cache it but I can't quite figure it out.

Devcontainer

• local variables (eg: {localEnv:USERNAME}) are not passed through properly from WSL2 to the container build

Dagger

• Does dagger contain its own internal docker engine?
• Can we use the hosts docker engine by default to avoid repeat image builds?