How mattwparas/steel works
Project type and purpose
• This is a monorepo for “Steel,” an embeddable Scheme dialect implemented in Rust.
• It provides:
• A parser and compiler pipeline that produces bytecode (and optional JIT),
• A VM engine exposed as a library (for embedding),
• A standalone CLI REPL,
• A language server (LSP) for editors,
• A standard library written in Scheme (the “cogs” directory),
• Documentation generation and book (mdBook),
• Examples, benchmarks, and tooling (xtask, Docker, Nix).
File structure
• Top-level directories:
• crates/ – Rust workspace crates (steel-core, steel-parser, steel-repl, steel-language-server, steel-doc, steel-derive, steel-gen, etc.)
• cogs/ – Scheme-level standard library modules and tests
• libs/ – Community/contrib libraries (markdown, regex, sqlite, HTTP, webserver, websockets, etc.)
• docs/ – mdBook documentation sources and theme
• examples/, steel-examples/, r7rs-benchmarks/, benchmarks/ – sample programs and performance tests
• src/ – likely legacy or top-level CLI glue (src/lib.rs, src/main.rs)
• Dockerfile, GitHub workflows, Nix files, Cargo.toml (workspace config), build.rs, xtask – build and CI tooling
Key insights from README
• Steel is R5RS-compliant with a powerful macro system (syntax-rules), contracts, transducers, and a bytecode VM with optional JIT.
• Exposes a Rust API (register_fn, register_type, run, extract).
• CLI REPL and an online playground exist.
• Language Server integrates with editors via LSP for diagnostics and auto-completion.
Be detailed but focused
• Every Rust crate or group them by function (parsing, core, repl, lsp, docs).
• How the cogs standard-library fits into module loading inside the VM.
• Embedding boundary clearly: anything calling Engine is an external consumer.