ajanibilby / bnf-parser Goto Github PK

View Code? Open in Web Editor NEW

9.0 2.0 1.0 2.31 MB

Deterministic BNF compiler/parser

Home Page: http://bnf-parser.ajanibilby.com/

License: MIT License

TypeScript 85.87% JavaScript 14.13%

bnf parser compiler javascript typescript

bnf-parser's Introduction

BNF-Parser

Compile your bnfs down to WebAssembly for maximum parsing speed; with generated type definitions to make using the syntax tree outputs a breeze. The compiled output from this library is platform agnostic, so it can run anywhere new WebAssembly.Instance() is a valid function. It bundles the WebAssembly module inside of a single js file so it can be easily incorporated into any bundler. The type definitions for a given bnf are just that, a definitions file - so if you don't want to use typescript or type hints you can go wild by ignoring it.

program ::= chunk+ ;
chunk ::= "a"+ "b"+ ;

npx bnf-compile ./syntax.bnf

import * as syntax from "./bnf/syntax.js";

const tree    = syntax.Parse_Program("abbaabab").root;
const chunk   = program.value[0]; // typescript knows this this **will** be of type `Term_Chunk`
const firstBs = program.value[1];
const bCount: number = firstBs.value.length; // typescript knows this **will** be a number

Built to be a devDependency - if you use the included cli tool to generate your syntax parser you don't need to include this library as your dependency, you can just import those artifacts.

Documentation

See https://bnf-parser.ajanibilby.com/

API

See https://bnf-parser.ajanibilby.com/api

BNF Syntax

See https://bnf-parser.ajanibilby.com/api

Try it Online

Try it in your browser https://bnf-parser.ajanibilby.com/test

bnf-parser's People

Contributors

Stargazers

Watchers

Forkers

pauldraper

bnf-parser's Issues

Not getting correct end bytes for repetitions of a select

Repetition of a select not working correctly:

# input: $abcHello $abc World
program ::= ("$abc" | !"$"+)+ ;

{
  "root": {
    "type": "program",
    "start": 0,
    "end": 0,
    "count": 1,
    "value": [
      {
        "type": "(...)+",
        "start": 0,
        "end": 0,
        "count": 4,
        "value": [
          {
            "type": "literal",
            "start": 0,
            "end": 4,
            "count": 4,
            "value": "$abc",
            "ref": {}
          }, {}, {},
          {
            "type": "literal",
            "start": 14,
            "end": 20,
            "count": 6,
            "value": " World",
            "ref": { }
          }
        ],
        "ref": {}
      }
    ],
    "ref": {},
  "reachBytes": 20,
  "reach": {
    "line": 1,
    "col": 21,
    "index": 20
  },
  "isPartial": true
}

This leads also to things being incorrectly labeled as partial, because reachBytes !== end

Repetition of sequences work correctly:

# input: $abcHello $abc World
program ::= ("$abc" !"$"+)+ ;

Example code is not working properly

+Show bnf stack trace on error

Include a list of strings within BNF_SyntaxError for the current BNF term that failed, and what terms it was "called" from. This can then be used to print out better syntax errors, allowing you to more easily debug BNFs due to the fact you can ensure that it is attempting to match the fail point with the correct term.

Hex codes don't match documentation

The documentation says that hex escapes work like "\x6b" but the actually behavior is "x6b".

Error with default BNF example

Hello,

Running through the homepage example, I use the following BNF:

program ::= %w* statement+ ;
block   ::= %("block" w+) ...name %(w* "{" w*) seq_stmt* %("}" w*) ;

And run:

npx bnf-compile ./syntax.bnf

And I get:

Found: ./syntax.bnf

 - Compiling syntax
   Encoding Parser
Parse Error: Unknown term name "w" (1:14) -> (1:16)

Cannot read properties of undefined (reading `'getReach'`) since v3.1.2

Hello!

bnf-parser 3.1.2 seems to have introduced a regression. We're getting the following crash when calling parse:

TypeError: Cannot read properties of undefined (reading 'getReach')
      at SyntaxNode.getReach (/Users/jeffrey.fowler/development/sequelize-v7-alias-bug/node_modules/bnf-parser/bin/syntax.js:70:50)
      at Sequence.parseSingle (/Users/jeffrey.fowler/development/sequelize-v7-alias-bug/node_modules/bnf-parser/bin/parser.js:347:32)
      at Sequence.parse (/Users/jeffrey.fowler/development/sequelize-v7-alias-bug/node_modules/bnf-parser/bin/parser.js:271:28)
      at Sequence.parse (/Users/jeffrey.fowler/development/sequelize-v7-alias-bug/node_modules/bnf-parser/bin/parser.js:323:25)
      at Rule.parse (/Users/jeffrey.fowler/development/sequelize-v7-alias-bug/node_modules/bnf-parser/bin/parser.js:389:28)
      at Parser.parse (/Users/jeffrey.fowler/development/sequelize-v7-alias-bug/node_modules/bnf-parser/bin/parser.js:425:29)
      at parseAttributeSyntaxInternal (/Users/jeffrey.fowler/development/sequelize-v7-alias-bug/node_modules/@sequelize/core/lib/utils/attribute-syntax.js:97:34)
      at memoized (/Users/jeffrey.fowler/development/sequelize-v7-alias-bug/node_modules/lodash/memoize.js:62:23)
      at /Users/jeffrey.fowler/development/sequelize-v7-alias-bug/node_modules/@sequelize/core/lib/dialects/abstract/where-sql-builder.js:215:88
      at Array.map (<anonymous>)

This is the BNF in question: https://github.com/sequelize/sequelize/blob/e6690cdf27980d57be1bb9f138b04124775b153b/packages/core/src/utils/attribute-syntax.ts#L56-L87
attributeParser.parse from the above code fails when we feed something like 'prop' into it.

Do you know if we're doing something incorrectly or if this is an actual bug?

Parse error due to select option order

// OK
let result = BNF.parse(`program ::= "aa" | "a";`);
let tree = Compile(result);
let syntax = tree.parse("aa");
console.log("result", result, "syntax", syntax);

// Not OK
let result = BNF.parse(`program ::= "a" | "aa";`);
let tree = Compile(result);
let syntax = tree.parse("aa");
console.log("result", result, "syntax", syntax);

Edit by @AjaniBilby:
Answer

Docs syntax link

The Syntax tab is broken for docs.

It should go to https://bnf-parser.ajanibilby.com/syntax but it doesn't.

Unparse

Very cool project!

What would be also nice is to have reverse parse/serialize functionality.

Should be possible without ambiguity (choose fewest possible repetitions).

Wasm-Parser: Upper case ranges not working

With the new wasm parser upper case ranges don't appear to work at all

This example will always fail

program ::= letter+ ;
letter ::= "a"->"z" | "A"->"Z" ;

snakeCase