Skip to content

Parse and stringify JSON with comments. It will retain comments even when after saved!

License

Notifications You must be signed in to change notification settings

kaelzhang/node-comment-json

Folders and files

NameName
Last commit message
Last commit date

Latest commit

Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 

Repository files navigation

Build Status Coverage npm module downloads per month

comment-json

Parse and stringify JSON with comments. It will retain comments even after saved!

  • Parse JSON strings with comments into JavaScript objects and MAINTAIN comments
    • supports comments everywhere, yes, EVERYWHERE in a JSON file, eventually πŸ˜†
    • fixes the known issue about comments inside arrays.
  • Stringify the objects into JSON strings with comments if there are

The usage of comment-json is exactly the same as the vanilla JSON object.

Table of Contents

Why?

There are many other libraries that can deal with JSON with comments, such as json5, or strip-json-comments, but none of them can stringify the parsed object and return back a JSON string the same as the original content.

Imagine that if the user settings are saved in ${library}.json, and the user has written a lot of comments to improve readability. If the library library need to modify the user setting, such as modifying some property values and adding new fields, and if the library uses json5 to read the settings, all comments will disappear after modified which will drive people insane.

So, if you want to parse a JSON string with comments, modify it, then save it back, comment-json is your must choice!

How?

comment-json parse JSON strings with comments and save comment tokens into symbol properties.

For JSON array with comments, comment-json extends the vanilla Array object into CommentArray whose instances could handle comments changes even after a comment array is modified.

Install

$ npm i comment-json

For TypeScript developers, @types/comment-json could be used

Since 2.4.1, comment-json contains typescript declarations, so you might as well remove @types/comment-json.

Usage

package.json:

{
  // package name
  "name": "comment-json"
}
const {
  parse,
  stringify,
  assign
} = require('comment-json')
const fs = require('fs')

const obj = parse(fs.readFileSync('package.json').toString())

console.log(obj.name) // comment-json

stringify(obj, null, 2)
// Will be the same as package.json, Oh yeah! πŸ˜†
// which will be very useful if we use a json file to store configurations.

Sort keys

It is a common use case to sort the keys of a JSON file

const parsed = parse(`{
  // b
  "b": 2,
  // a
  "a": 1
}`)

// Copy the properties including comments from `parsed` to the new object `{}`
// according to the sequence of the given keys
const sorted = assign(
  {},
  parsed,
  // You could also use your custom sorting function
  Object.keys(parsed).sort()
)

console.log(stringify(sorted, null, 2))
// {
//   // a
//   "a": 1,
//   // b
//   "b": 2
// }

For details about assign, see here.

parse()

parse(text, reviver? = null, remove_comments? = false)
  : object | string | number | boolean | null
  • text string The string to parse as JSON. See the JSON object for a description of JSON syntax.
  • reviver? Function() | null Default to null. It acts the same as the second parameter of JSON.parse. If a function, prescribes how the value originally produced by parsing is transformed, before being returned.
  • remove_comments? boolean = false If true, the comments won't be maintained, which is often used when we want to get a clean object.

Returns CommentJSONValue (object | string | number | boolean | null) corresponding to the given JSON text.

If the content is:

/**
 before-all
 */
// before-all
{ // before:foo
  // before:foo
  /* before:foo */
  "foo" /* after-prop:foo */: // after-colon:foo
  1 // after-value:foo
  // after-value:foo
  , // after:foo
  // before:bar
  "bar": [ // before:0
    // before:0
    "baz" // after-value:0
    // after-value:0
    , // after:0
    "quux"
    // after:1
  ] // after:bar
  // after:bar
}
// after-all
const {inspect} = require('util')

const parsed = parse(content)

console.log(
  inspect(parsed, {
    // Since 4.0.0, symbol properties of comments are not enumerable,
    // use `showHidden: true` to print them
    showHidden: true
  })
)

console.log(Object.keys(parsed))
// > ['foo', 'bar']

console.log(stringify(parsed, null, 2))
// πŸš€ Exact as the content above! πŸš€

And the value of parsed will be:

{
  // Comments before the JSON object
  [Symbol.for('before-all')]: [{
    type: 'BlockComment',
    value: '\n before-all\n ',
    inline: false,
    loc: {
      // The start location of `/**`
      start: {
        line: 1,
        column: 0
      },
      // The end location of `*/`
      end: {
        line: 3,
        column: 3
      }
    }
  }, {
    type: 'LineComment',
    value: ' before-all',
    inline: false,
    loc: ...
  }],
  ...

  [Symbol.for('after-prop:foo')]: [{
    type: 'BlockComment',
    value: ' after-prop:foo ',
    inline: true,
    loc: ...
  }],

  // The real value
  foo: 1,
  bar: [
    "baz",
    "quux",

    // The property of the array
    [Symbol.for('after-value:0')]: [{
      type: 'LineComment',
      value: ' after-value:0',
      inline: true,
    loc: ...
    }, ...],
    ...
  ]
}

There are EIGHT kinds of symbol properties:

// Comments before everything
Symbol.for('before-all')

// If all things inside an object or an array are comments
Symbol.for('before')

// comment tokens before
// - a property of an object
// - an item of an array
// and after the previous comma(`,`) or the opening bracket(`{` or `[`)
Symbol.for(`before:${prop}`)

// comment tokens after property key `prop` and before colon(`:`)
Symbol.for(`after-prop:${prop}`)

// comment tokens after the colon(`:`) of property `prop` and before property value
Symbol.for(`after-colon:${prop}`)

// comment tokens after
// - the value of property `prop` inside an object
// - the item of index `prop` inside an array
// and before the next key-value/item delimiter(`,`)
// or the closing bracket(`}` or `]`)
Symbol.for(`after-value:${prop}`)

// comment tokens after
// - comma(`,`)
// - the value of property `prop` if it is the last property
Symbol.for(`after:${prop}`)

// Comments after everything
Symbol.for('after-all')

And the value of each symbol property is an array of CommentToken

interface CommentToken {
  type: 'BlockComment' | 'LineComment'
  // The content of the comment, including whitespaces and line breaks
  value: string
  // If the start location is the same line as the previous token,
  // then `inline` is `true`
  inline: boolean

  // But pay attention that,
  // locations will NOT be maintained when stringified
  loc: CommentLocation
}

interface CommentLocation {
  // The start location begins at the `//` or `/*` symbol
  start: Location
  // The end location of multi-line comment ends at the `*/` symbol
  end: Location
}

interface Location {
  line: number
  column: number
}

Query comments in TypeScript

comment-json provides a symbol-type called CommentSymbol which can be used for querying comments. Furthermore, a type CommentDescriptor is provided for enforcing properly formatted symbol names:

import {
  CommentDescriptor, CommentSymbol, parse, CommentArray
} from 'comment-json'

const parsed = parse(`{ /* test */ "foo": "bar" }`)
 // typescript only allows properly formatted symbol names here
const symbolName: CommentDescriptor = 'before:foo'

console.log((parsed as CommentArray<string>)[Symbol.for(symbolName) as CommentSymbol][0].value)

In this example, casting to Symbol.for(symbolName) to CommentSymbol is mandatory. Otherwise, TypeScript won't detect that you're trying to query comments.

Parse into an object without comments

console.log(parse(content, null, true))

And the result will be:

{
  foo: 1,
  bar: [
    "baz",
    "quux"
  ]
}

Special cases

const parsed = parse(`
// comment
1
`)

console.log(parsed === 1)
// false

If we parse a JSON of primative type with remove_comments:false, then the return value of parse() will be of object type.

The value of parsed is equivalent to:

const parsed = new Number(1)

parsed[Symbol.for('before-all')] = [{
  type: 'LineComment',
  value: ' comment',
  inline: false,
  loc: ...
}]

Which is similar for:

  • Boolean type
  • String type

For example

const parsed = parse(`
"foo" /* comment */
`)

Which is equivalent to

const parsed = new String('foo')

parsed[Symbol.for('after-all')] = [{
  type: 'BlockComment',
  value: ' comment ',
  inline: true,
  loc: ...
}]

But there is one exception:

const parsed = parse(`
// comment
null
`)

console.log(parsed === null) // true

stringify()

stringify(object: any, replacer?, space?): string

The arguments are the same as the vanilla JSON.stringify.

And it does the similar thing as the vanilla one, but also deal with extra properties and convert them into comments.

console.log(stringify(parsed, null, 2))
// Exactly the same as `content`

space

If space is not specified, or the space is an empty string, the result of stringify() will have no comments.

For the case above:

console.log(stringify(result)) // {"a":1}
console.log(stringify(result, null, 2)) // is the same as `code`

assign(target: object, source?: object, keys?: Array)

  • target object the target object
  • source? object the source object. This parameter is optional but it is silly to not pass this argument.
  • keys? Array<string> If not specified, all enumerable own properties of source will be used.

This method is used to copy the enumerable own properties and their corresponding comment symbol properties to the target object.

const parsed = parse(`// before all
{
  // This is a comment
  "foo": "bar"
}`)

const obj = assign({
  bar: 'baz'
}, parsed)

stringify(obj, null, 2)
// // before all
// {
//   "bar": "baz",
//   // This is a comment
//   "foo": "bar"
// }

Special cases about keys

But if argument keys is specified and is not empty, then comment before all, which belongs to no properties, will NOT be copied.

const obj = assign({
  bar: 'baz'
}, parsed, ['foo'])

stringify(obj, null, 2)
// {
//   "bar": "baz",
//   // This is a comment
//   "foo": "bar"
// }

Specifying the argument keys as an empty array indicates that it will only copy non-property symbols of comments

const obj = assign({
  bar: 'baz'
}, parsed, [])

stringify(obj, null, 2)
// // before all
// {
//   "bar": "baz",
// }

Non-property symbols include:

Symbol.for('before-all')
Symbol.for('before')
Symbol.for('after-all')

CommentArray

Advanced Section

All arrays of the parsed object are CommentArrays.

The constructor of CommentArray could be accessed by:

const {CommentArray} = require('comment-json')

If we modify a comment array, its comment symbol properties could be handled automatically.

const parsed = parse(`{
  "foo": [
    // bar
    "bar",
    // baz,
    "baz"
  ]
}`)

parsed.foo.unshift('qux')

stringify(parsed, null, 2)
// {
//   "foo": [
//     "qux",
//     // bar
//     "bar",
//     // baz
//     "baz"
//   ]
// }

Oh yeah! πŸ˜†

But pay attention, if you reassign the property of a comment array with a normal array, all comments will be gone:

parsed.foo = ['quux'].concat(parsed.foo)
stringify(parsed, null, 2)
// {
//   "foo": [
//     "quux",
//     "qux",
//     "bar",
//     "baz"
//   ]
// }

// Whoooops!! 😩 Comments are gone

Instead, we should:

parsed.foo = new CommentArray('quux').concat(parsed.foo)
stringify(parsed, null, 2)
// {
//   "foo": [
//     "quux",
//     "qux",
//     // bar
//     "bar",
//     // baz
//     "baz"
//   ]
// }

Special Cases about Trailing Comma

If we have a JSON string str

{
  "foo": "bar", // comment
}
// When stringify, trailing commas will be eliminated
const stringified = stringify(parse(str), null, 2)
console.log(stringified)

And it will print:

{
  "foo": "bar" // comment
}

License

MIT

Change Logs

See releases