8.5 KiB

Raw Permalink Blame History

json-ext

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

Features:

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

Install

npm install @discoveryjs/json-ext

API

parseChunked(chunkEmitter)
stringifyStream(value[, replacer[, space]])
stringifyInfo(value[, replacer[, space[, options]]])
- Options
  - async
  - continueOnCircular
version

parseChunked(chunkEmitter)

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.

Benchmark

Usage:

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

// as a regular Promise
parseChunked(chunkEmitter)
    .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:

ReadableStream (Node.js only)

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

parseChunked(fs.createReadStream('path/to/file.json'))

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) {
                break;
            }

            yield value;
        }
    });
}

loadData('https://example.com/data.json')
    .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)

Benchmark

Usage:

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

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

// pipe into a stream
stringifyStream(data)
    .pipe(writableStream);

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!"}}}
stringifyStream({
    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
stringifyStream({
    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
stringifyStream(data)
    .pipe(process.stdout);

// pipe into a file
stringifyStream(data)
    .pipe(fs.createWriteStream('path/to/file.json'));

// wrapping into a Promise
new Promise((resolve, reject) => {
    stringifyStream(data)
        .on('error', reject)
        .pipe(stream)
        .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
}

Example:

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

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

Options

async

Type: Boolean
Default: false

Collect async values (promises and streams) or not.

continueOnCircular