BPMN-based creation of SimPy discrete event simulation models

Wouldn't it be cool to combine the block-based process modeling experience of commercial discrete event simulation packages with the amenities of proper IDE-based source-code editing? (Think Arena / Anylogic / ExtendSim / Plant Simulation / ... but with simple integration of third-party libraries, industry-standard interfaces, unit- and integration testing, dockerization, serverless execution in the cloud of your choice... and even actually working auto-completion! :D)

And all that not only for free, but using the worlds most popular language for data analytics and machine learning?

Casymda enables you to create SimPy simulation models, with help of BPMN and the battle-tested Camunda-Modeler.

Created BPMN process diagrams are parsed and converted to Python-code, combining visual oversight of model structure with code-based definition of model behavior. Immediately executable, including a token-based process animation, allowing for space-discrete entity movements, and ready to be wrapped as a gym-environment to let a machine-learning algorithm find a control strategy.

Further information and sample projects:

• https://fladdimir.github.io/

Installation

From PyPI:

pip install casymda

Features

• connectable blocks for processing of entities
• graphical model description via camunda modeler
• process visualization browser-based or via tkinter
• space-discrete tilemap-movements of entities
• gradually typed

Coming soon:

• automated model generation from process event-logs via PM4Py

Examples

Basic features are illustrated as part of the example models (which also serve as integration tests):

• basics:
  • bpmn-based generation of a simple model file
  • run the generated model
  • process visualization via tkinter
  • browser-based visualization (served with flask, animated with pixijs)
• resources:
  • seize and release a resource via graphical modeling
• tilemap:
  • entity movement along a shortest path defined by a csv-tilemap (built on networkx: https://networkx.github.io/)

For setup just clone the repository and install casymda (virtual environment recommended). See basics-visual-run-tkinter for an example of how to cope with python-path issues.

Design

• Model generation and execution
• Blocks and animation

Development

This project trusts Black for formatting, Sonarqube for static code analysis, and pytest for unit & integration testing. Developed and tested on Linux (Ubuntu 20.04), Python 3.8.5. Tests can be carried out inside a docker-container, optionally including an installation from pypi to verify a successful upload.

Sonarqube

sonarqube server (public docker image):

docker-compose up sonarqube

sonar-scanner (public docker image):

docker-compose up analysis

(run a docker-based unit-test first for coverage-reporting)
(remember to share your drive via Docker-Desktop settings if necessary, to be re-applied after each password change)

Tests

pytest --cov-report term --cov=src/casymda/ tests/

integrations tests:

python3 -m pytest examples

(integration-tests require tkinter, which may be installed via sudo apt-get install python3-tk)

For Docker-based tests see docker-compose.yml

docker-compose run unit-test
docker-compose run examples-test
docker-compose run examples-test-pypi

Virtual environment setup

python3 -m venv venv

Editable installation

pip install -e .

Publish to pypi

python setup.py sdist

twine upload dist/*

pip install twine if necessary,
remember to set the version in setup.py and src/casymda as required

PyPy3

Tested PyPy3 (7.3.1-final). Install pypy3 pypy3-dev pypy-tk.

Runtime could be decreased by factor ~2 when using PyPy3 for longer simulations runs (e.g. from ~45s to ~25s for a simple example model test with MAX_ENTITIES set to 40.000 on an i5 notebook).