This repository is a mono-repo containing the following parts: - process_mining/ Process Mining library in Rust (See also https://crates.io/crates/processmining) - `pythonbridge/Python Bindings -javabridge/Java Bindings -pmwasm/WASM / JavaScript Bindings PoC -binary/` Rust Binary Executable

For more information on the process_mining Rust library, see process_mining/README.md. \ Further information is also available at https://crates.io/crates/processmining and https://docs.rs/processmining.

Setup and Project Structure

The Java, Python and JavaScript bindings make use of the main process_mining library, which is written in Rust. Thus, complex algorithms or logic can be implemented only once in Rust. This implementation can then be exposed and used from Java, Python or JavaScript, in different enviroments. This not only reduces implementation effort (write only once, use everywhere), but also enables great performance.

Additional information, as well as evaluation results, are published in a written report titled Developing a High-Performance Process Mining Library with Java and Python Bindings in Rust.

Mostly, cargo is used to build and run this project. For more detailed instructions see https://github.com/aarkue/rust-bridge-template, which provides more details on how to build or run this and similar projects.

If you are looking for the Rust-based XES event log importer, see https://github.com/aarkue/rustxes, which contains the XES importer from process_mining/ extracted to its own crate with full Python bindings.

If you are interested in a starter kit template to bootstrap similar Rust libraries with Java and Python bindings, see https://github.com/aarkue/rust-bridge-template.

There you will find more information on how to get started, as well as build and run such projects.

Main Rust Library (process_mining)

process_mining is a collection of functions, structs and utilitities related to Process Mining.

See process_mining/README.md to view the dedicated README for the process_mining Rust library.

Further information is also available at https://crates.io/crates/processmining and https://docs.rs/processmining.

Bindings

Here you can find some preliminary information about the binding implementations. For more details, please take a look at the source code or the full written report.

Java

• Uses jni to allow Java code to call shared Rust library
• Object Passing:
  • byte[] JSON-encoded data [faster]
  • Writing File to disk (containing JSON-encoded data)
• Required Java libraries:
  • gson for efficient JSON encoding/decoding

Python

• Uses maturin with the corresponding PyO3 FFI Bindings
• Can pass dicts and other types rather easily; but slow for larger data
• JSON-encoding/decoding possible (using faster orjson), but still no great passing performance
  • bytes encoding/decoding e.g., Event Logs is implemented__
• Polars: Allows converting PM4Py's pandas DataFrame to a Polars DataFrame; Can then easily be used on Rust side
  • But: Requires Polars dependency (also in python!)

WASM Proof of Concept

• pm_wasm/ contains a proof of concept of WASM bindings for the main library, e.g., exposed to JavaScript and executed in the browser.
• First building with wasm-pack build --target web --release
• The PoC frontend uses React with vite. The dev server can be started with npm run dev.
• Multithreading (using wasm-bindgen-rayon) is currently not used, but might be enabled again in the future.

A live demo is available at https://wasm.siter.eu/.

Example XES -> Alpha+++ Discovery | Example OCEL2.0 XML Import :-------------------------:|:-------------------------: |

LICENSE

This project is licensed under either Apache License Version 2.0 or MIT License at your option.