A set of utilities that extend the use of JSON

Usage no npm install needed!

<script type="module">
  import discoveryjsJsonExt from 'https://cdn.skypack.dev/@discoveryjs/json-ext';



NPM version Build Status Coverage Status

A set of utilities that extend the use of JSON. Designed to be fast and memory efficient


  • parseChunked() – Parse JSON that comes by chunks (e.g. FS readable stream or fetch response stream)
  • stringifyStream() – Stringify stream (Node.js)
  • stringifyInfo() – Get estimated size and other facts of JSON.stringify() without converting a value to string
  • TBD Support for circular references
  • TBD Binary representation branch
  • TBD WHATWG Streams support


npm install @discoveryjs/json-ext



Works the same as JSON.parse() but takes chunkEmitter instead of string and returns Promise.

NOTE: reviver parameter is not supported yet, but will be added in next releases. NOTE: WHATWG streams aren't supported yet

When to use:

  • It's required to avoid freezing the main thread during big JSON parsing, since this process can be distributed in time
  • Huge JSON needs to be parsed (e.g. >500MB on Node.js)
  • Needed to reduce memory pressure. JSON.parse() needs to receive the entire JSON before parsing it. With parseChunked() you may parse JSON as first bytes of it comes. This approach helps to avoid storing a huge string in the memory at a single time point and following GC.



const { parseChunked } = require('@discoveryjs/json-ext');

// as a regular Promise
    .then(data => {
        /* data is parsed JSON */

// using await (keep in mind that not every runtime has a support for top level await)
const data = await parseChunked(chunkEmitter);

Parameter chunkEmitter can be:

const fs = require('fs');
const { parseChunked } = require('@discoveryjs/json-ext');

  • Generator, async generator or function that returns iterable (chunks). Chunk might be a string, Uint8Array or Buffer (Node.js only):
const { parseChunked } = require('@discoveryjs/json-ext');
const encoder = new TextEncoder();

// generator
parseChunked(function*() {
    yield '{ "hello":';
    yield Buffer.from(' "wor');    // Node.js only
    yield encoder.encode('ld" }'); // returns Uint8Array(5) [ 108, 100, 34, 32, 125 ]

// async generator
parseChunked(async function*() {
    for await (const chunk of someAsyncSource) {
        yield chunk;

// function that returns iterable
parseChunked(() => ['{ "hello":', ' "world"}'])

Using with fetch():

async function loadData(url) {
    const response = await fetch(url);
    const reader = response.body.getReader();

    return parseChunked(async function*() {
        while (true) {
            const { done, value } = await reader.read();

            if (done) {

            yield value;

    .then(data => {
        /* data is parsed JSON */

stringifyStream(value[, replacer[, space]])

Works the same as JSON.stringify(), but returns an instance of ReadableStream instead of string.

NOTE: WHATWG Streams aren't supported yet, so function available for Node.js only for now

Departs from JSON.stringify():

  • Outputs null when JSON.stringify() returns undefined (since streams may not emit undefined)
  • A promise is resolving and the resulting value is stringifying as a regular one
  • A stream in non-object mode is piping to output as is
  • A stream in object mode is piping to output as an array of objects

When to use:

  • Huge JSON needs to be generated (e.g. >500MB on Node.js)
  • Needed to reduce memory pressure. JSON.stringify() needs to generate the entire JSON before send or write it to somewhere. With stringifyStream() you may send a result to somewhere as first bytes of the result appears. This approach helps to avoid storing a huge string in the memory at a single time point.
  • The object being serialized contains Promises or Streams (see Usage for examples)



const { stringifyStream } = require('@discoveryjs/json-ext');

// handle events
    .on('data', chunk => console.log(chunk))
    .on('error', error => consold.error(error))
    .on('finish', () => console.log('DONE!'));

// pipe into a stream

Using Promise or ReadableStream in serializing object:

const fs = require('fs');
const { stringifyStream } = require('@discoveryjs/json-ext');

// output will be
// {"name":"example","willSerializeResolvedValue":42,"fromFile":[1, 2, 3],"at":{"any":{"level":"promise!"}}}
    name: 'example',
    willSerializeResolvedValue: Promise.resolve(42),
    fromFile: fs.createReadStream('path/to/file.json'), // support file content is "[1, 2, 3]", it'll be inserted as it
    at: {
        any: {
            level: new Promise(resolve => setTimeout(() => resolve('promise!'), 100))

// in case several async requests are used in object, it's prefered
// to put fastest requests first, because in this case
    foo: fetch('http://example.com/request_takes_2s').then(req => req.json()),
    bar: fetch('http://example.com/request_takes_5s').then(req => req.json())

Using with WritableStream (Node.js only):

const fs = require('fs');
const { stringifyStream } = require('@discoveryjs/json-ext');

// pipe into a console

// pipe into a file

// wrapping into a Promise
new Promise((resolve, reject) => {
        .on('error', reject)
        .on('error', reject)
        .on('finish', resolve);

stringifyInfo(value[, replacer[, space[, options]]])

value, replacer and space arguments are the same as for JSON.stringify().

Result is an object:

    minLength: Number,  // minimal bytes when values is stringified
    circular: [...],    // list of circular references
    duplicate: [...],   // list of objects that occur more than once
    async: [...]        // list of async values, i.e. promises and streams


const { stringifyInfo } = require('@discoveryjs/json-ext');

    stringifyInfo({ test: true }).minLength
// > 13
// that equals '{"test":true}'.length



Type: Boolean
Default: false

Collect async values (promises and streams) or not.


Type: Boolean
Default: false

Stop collecting info for a value or not whenever circular reference is found. Setting option to true allows to find all circular references.


The version of library, e.g. "0.3.1".