dsel-glr-parser

A small, dependency-free GLR (generalized LR) parser for the DSEL language described in the prompt:

• S-expression-ish lists with ( / )
• reserved head tokens: ->, id->
• labels: bare or backticked (with escapes)
• arrows are first-class terms (arrow endpoints can be arrows)
• parenthesis mismatch recovery (drops extra ) and inserts missing ) at EOF)

It produces the minimal AST form:

export type Thing =
  | { kind: "atom"; id: string; label: string }
  | { kind: "box"; id: string; label: string; items: Thing[] }
  | { kind: "arrow"; id: string; label: string | null; from: Thing; to: Thing };

(Self-loops are arrows with from === to.)

Install

npm i dsel-glr-parser

Usage

import { parse } from "dsel-glr-parser";

const src = `(-> is_a square rectangle)`;
const res = parse(src, { mode: "recover" });

console.log(res.ast);
console.log(res.diagnostics);

API

• tokenize(input: string): { tokens: Token[]; diagnostics: Diagnostic[] }
• parse(input: string, options?): ParseResult

Options:

• mode: "strict" throws on diagnostics that make the document invalid; "recover" returns best-effort AST.
• recoverParentheses: default true
• maxRecoveries: default 50 (limits skip/insert recovery actions)

Development

npm i
npm test
npm run build

Notes on recovery

This library explicitly focuses on mismatched parenthesis recovery:

• If there are extra ), they are dropped (diagnostic emitted).
• If there are missing ), they are inserted at EOF (diagnostic emitted).

For other syntax issues (like invalid arrow arity), in mode: "recover" it emits diagnostics and produces a best-effort AST by:

• inserting placeholder atoms labeled "<missing>" when required,
• ignoring extra arguments.

In mode: "strict", these are errors.