An implementation of the SWIRLDS HashGraph Consensus Algorithm as specified in the paper. This is a library that provides all the functions needed to reach consensus.
This code uses iterators to traverse commitments, and although it is a pretty abstraction, it can be slow. I re-implemented the library using conventional graph search algorithms and matrix math here. Performance is improved dramatically.
The library compiles to WebAssembly and can be
invoked like a normal ES6 module. The www directory contains a javascript
webapp to interactively build and visualize a hashgraph as a force directed
graph in D3.js.
Right now tests aren't work as the interface has changed to support wasm
Run the tests with cargo test. The unit tests in the graph module are the necessary steps to
order all events topologically according to the HashGraph algorithm.
The main.rs file contains the project binary. This is effectively an event loop that will
- Connect and discover peers over the mDNS protocol
- Upon receiving an event from a peer, update the in-memory graph, and pass the event onto another random peer
- Host an HTTP service that a user can upload new transactions to, which will be encoded into an event and passed to a random peer
- Nodes send and receive events on gossip network with rust-libp2p
- Graph traversal optimizations
- Right now cloning strings is abundant to support the js/wasm cross-boundary. There should be workarounds for this, or at least a macro system to generate an optimized version for native.