MORK.jl

A Julia implementation of the MORK metagraph rewriting engine — the high-performance, trie-native substrate for symbolic computation, built on PathMap.jl.

MORK is an independent Julia port of the Rust MORK kernel, hardened by a deep Rust→Julia porting audit (intern races, numeric-primitive width, COW discipline, operator-precedence transpilation defects).

Features

  • Expression engine — byte-encoded s-expressions over a PathMap{UnitVal} substrate, with prefix-scoped multi-space queries.
  • Transform / exec — pattern→template rewriting, sinks (Count/Head/AU/…), and the metta-calculus step.
  • Concurrent symbol interning — thread-safe bucket-map symbol table.
  • Numeric primitives — full u8…u128 / i8…i128 / f32 / f64 op suite.

Worked example: the fly connectome

A full real-world build — the FAFB v783 Drosophila connectome (≈139 k neurons, 3.73 M synapses) organized as persisted per-Space .act snapshots, with the Nature 2024 Fig-6 information-flow model run over them: Fly connectome (FAFB v783).

It showcases the substrate end-to-end: prefix-scoped multi-Space queries, cold-mmap .act reads (~0.25 ms open of a 41.7 MB snapshot), and the zipper-frontier traversal from the Zipper Queries guide — validated node-for-node against an independent oracle.

Installation

using Pkg
Pkg.add(url="https://github.com/CognitiveSubstratesAI/MORK")

Contents