Unable to reverse

[mayask]

Hey

I'm trying to use nearley-reverse to be able to generate back original text after parsing using the same grammar definition, but I get really weird result.

# mt101.nearley.ne

@{%
function factory(decode, encode) {
  decode.encode = encode
  return decode
}

id.encode = x => [x]
%}

genericRecord -> ":" recordId ":" recordContent
    {% factory(
        ([_1, recordId, _2, content]) => {
            return {
                recordId,
                content,
            }
        },
        ([ _1, recordId, _2, content ]) => {
            return [":", recordId, ":", content]
        }
    ) %}

recordId -> [0-9A-Z]:+
    {% factory(
        ([d]) => d.join(''),
        rid => (rid || '').split(''),
    ) %}

recordContent -> [^:]:+
    {% factory(
        ([d]) => d.join(''),
        c => (c || '').split(''),
    ) %}

// test.js

let nearley = require('nearley')
let reverse = require('nearley-reverse')
let compiledGrammar = require('./mt101.nearley')

let mt101Text = `:20:1106210100000003`

let grammar = nearley.Grammar.fromCompiled(compiledGrammar)
let parser = new nearley.Parser(grammar)

parser.feed(mt101Text)

let result = parser.results[0]

console.log('json\n', JSON.stringify(result, null, 2))
console.log()

let original = reverse(grammar, [':', '20', ':', '1106210100000003']).join('')
console.log('original\n', original)

# output

json
 {
  "recordId": "20",
  "content": "1106210100000003"
}

original
 :/[0-9A-Z]/:/[^:]/

What am I doing wrong?

[tjvr]

Your encoders aren't quite right. They should take a piece of parse tree and return a list, if I remember correctly:

x => {
            return [":", x.recordId, ":", x.content]
        }

Finally, I don't know if nearley-reverse supports character groups like [0-9A-Z]; but it definitely doesn't like EBNF modifiers like :+.

I hope that helps! 🙂

[mayask]

Hi @tjvr Got it. Is there a way to handle ebnf modifiers? Do I have to use tokenizer for that?

[mayask]

Any thoughts @tjvr ?

[tjvr]

I think nearley-reverse is supposed to be used with a tokenizer, yes.

If you really need EBNF modifiers, you can always write them yourself, e.g. A -> B:+ becomes

A -> B
      | A B

[mayask]

I just replaced one field that was previously using EBNF modifiers with a custom rule for :+:

recordIdChar -> [0-9A-Z]

recordId -> recordIdChar
    | recordId recordIdChar

And I still get the same result