@tabnas/ebnf

EBNF grammar compiler for the Tabnas parser: takes EBNF source in the W3C dialect and emits a tabnas GrammarSpec.

This is a best-effort front-end. It implements one dialect properly and refuses the rest by name: ISO/IEC 14977 subtraction (A - B), special sequences (? … ?) and bracket repetition ({ A }) all raise a named error rather than compiling to something plausible. The ISO option [ A ] is the one to watch, because [ … ] is a character class in this dialect: write A?. The itemised list is in doc/reference.md; read it before writing a grammar.

Postfix operators may be stacked, but every unbounded * or + layer must consume input on each iteration. A form such as (A?)* is refused at compile time instead of emitting a loop that can stay at one input position forever.

Install

npm install @tabnas/parser @tabnas/bnf @tabnas/ebnf

@tabnas/parser (the engine) and @tabnas/bnf (the shared compiler) are peer dependencies.

One example

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

const tn = new Tabnas({ plugins: [ebnf] })
tn.ebnf(`
  Expr   ::= Term ( ( "+" | "-" ) Term )*
  Term   ::= Factor ( ( "*" | "/" ) Factor )*
  Factor ::= NR | "(" Expr ")"
`)

tn.parse('1 + 2 * 3').src // => '1+2*3'
tn.parse('(1+2)*3').rule  // => 'Expr'

Build the instance once and reuse it: compiling the grammar is the expensive part. tn.ebnf.toSpec(src) builds the spec without installing it.

Documentation

Four-quadrant Diátaxis docs:

| Quadrant | File | |---|---| | Tutorial (learn) | doc/tutorial.md | | How-to guide (tasks) | doc/guide.md | | Reference (API + dialect) | doc/reference.md | | Concepts (why) | doc/concepts.md |

Repo-level orientation, including the full supported/unsupported table, is in the top-level README.

License

MIT. Copyright (c) Richard Rodger.