ABNF grammar test bench

What it does

Run reusable pass/fail sample suites against an ABNF grammar, with failure offsets and successful rule spans.

How to use it

1. Paste ABNF, choose a start rule and supply a JSON array of input strings and expected match booleans.
2. Run the suite, inspect diagnostics and export the results or a reloadable fixture bundle. Core rules and RFC 7405 strings are supported.
3. Processing stays in a bounded local worker. No generated code, external grammar resolution or upload.

Example

The command grammar accepts "GET /docs/start\r\n" and rejects lowercase "get" because the method uses a case-sensitive %s literal.

Methodology

ABNF is compiled into a bounded context-free grammar and recognized with an Earley chart, supporting recursion, alternatives, repetition and nullable rules. Core rules follow RFC 5234; %s/%i follow RFC 7405. Diagnostics show farthest reachable code-point offset and expected terminals; a match includes one derivation, not every ambiguous parse.

Limitations

50,000 grammar characters, 500 definitions, nesting 30, explicit repeat bounds 256, 5,000 productions. Up to 50 samples / 2,048 code points each / 20,000 characters total; 100,000 states per sample, five million operations per batch and 30 seconds. Prose values need explicit replacement rules. Input is Unicode code points, not encoded bytes. LF/CRLF and indented definitions are accepted. No protocol conformance certification or remote grammar loading.

Automation

This tool runs in your browser. No upload API or MCP endpoint is available.