tabnas

A pluggable parsing engine. The runtime is a class (Tabnas) that runs a rule-based parser over a configurable matcher-based lexer. The package ships no grammar of its own: every grammar is a plugin that you (or another package) supply.

npm install @tabnas/parser

A tiny taste, a one-token grammar defined inline:

const { Tabnas } = require('@tabnas/parser')

const tn = new Tabnas({ plugins: [(tn) => {
  tn.options({ fixed: { token: { '#HI': 'hello' } } })
  tn.rule('val', (rs) => rs.open([
    { s: ['#HI'], a: (r) => { r.node = 'world' } },
  ]))
}] })

tn.parse('hello')                     // 'world'

That grammar as a railroad/syntax diagram, generated from the live parser with @tabnas/railroad:

A vertical ASCII version is in doc/taste.txt.

Documentation

• Tutorial. Build your first parser, step by step.
• How-to guides. Short recipes for common tasks.
• Writing plugins. Author a grammar or tooling plugin.
• API reference. Every public method, property, and export.
• Options reference. Every option field and its default.
• Concepts. How the TypeScript engine is put together, and why.

Shared design docs for both runtimes live at the top of the repo:

• Architecture. The engine model.
• Syntax. The relaxed-JSON syntax reference.

Explanation / design notes

• LSP feasibility. Language-server angles.
• GBNF feasibility. Llama.cpp grammar (constrained LLM decoding) angles.

Companion packages

Grammars and tooling ship as separate packages, not in this one:

• @tabnas/abnf. Compiles ABNF into engine rules and adds tn.bnf(src).
• @tabnas/debug. Tracing and describe() helpers.

The strict-JSON grammar used by the conformance tests lives as a test fixture at test/json-plugin.ts: a worked example of a non-trivial grammar plugin.

The Go port follows the same grammar-free design, with its strict-JSON grammar kept as a test fixture too.

License

MIT. Copyright (c) Richard Rodger.