SJS

The fastest JSON stringifier

TOC

• SJS
  • TOC
  • TL;DR
  • Installation
  • How it works
  • Caveats
  • Benchmarks
    • Running the benchmarks
      • Test machine
      • Some numbers
  • Usage
    • Supported types
      • Defining a schema
      • Defining schema with simple array
      • Defining schema with complex array
      • Defining schema with nested objects
  • Serializers
  • API
    • sjs
    • attr
    • escape
  • License

TL;DR

SJS shows a significant increase in performance over both native JSON.stringify and fast-json-stringify. For some use cases (dealing with long text), it performs 21000% faster than both native and fast-json-stringify.

NOTE: Support for undefined properties has been added from 1.0.1. SJS is now production ready.

Checkout benchmarks.

Installation

SJS is fully compatible with both Node.js and the browser 🎉🎉

Node:

1npm install slow-json-stringify

On the browser:

1<script src="https://unpkg.com/slow-json-stringify/dist/sjs.umd.js"></script>

How it works

Why SJS is the fastest stringifier? The traditional approach consists in serializing every property taken singularly. SJS uses a different approach to serialization.

Preparation:

• A schema is provided
• The schema is stringified
• A templated string is built with the provided schema.

Serialization:

• Object values are inserted in the already built template.

It is faster simply because it performs a lot less work.

Caveats

SJS require some setup work if compared to native JSON.stringify. But, if you are dealing with json with a fixed structure SJS will save you a ton of time. Especially when the payload grows. And incredibly when serializing json with long text inside (think of a blog article or a product description. Moreover, SJS makes possible the serialization of types that are not natively supported by JSON.stringify thanks to custom serializers.

note: SJS won't perform any escaping as you usually won't need it in small payloads. If you are working with big text, it could be of very little effort to store in your db an already escaped text.

However, SJS provides a little utility for your escaping needs. escape uses a default regex if no additional regex is provided.

default regex string:

1/\n|\r|\t|\"|\\/gm

You can use escape like the following:

1const { escape } = require('slow-json-stringify');
2
3// If you don't pass any additional regex, a default one will be used.
4const escaper = escape();
5
6escaper('This is "funny"'); // This is \"funny\"
7
8// You can pass any regex you want for your escaping strategy.
9const customEscaper = escape(/\"/gm);
10
11customEscaper('This is "funny"'); // This is \"funny\"
12

Benchmarks

We all know that there are three kinds of lies..

Lies, damned lies.. and benchmarks.

Remember to test if SJS could be a real improvement for your use case. Because there are times when the performance advantages with the added drawbacks could not be worth it.

Running the benchmarks

Every benchmark is replicable on your own machine. To run your tests:

• Clone this repo.
• Install dependencies.
• cd benchmark.
• Grant executable rights to run.sh script chmod +x ./run.sh.
• Save benchmark results to file ./run.sh >> benchmark.md

Test machine

The benchmarks were performed on a Dell Xps 15 9550.

• cpu: intel i7 6700HQ
• ram: 16gb
• os: Ubuntu 18.04

Some numbers

Checkout benchmarks here

Usage

Supported types

The schema creation happens thanks to the attr helper exported from the main bundle.

1const { attr } = require('sjs');

1attr(type, serializer?)

The attr helper natively supports the following types:

• string
• number
• boolean
• null
• array, dynamic array with simple structure, in this scenario native JSON.stringify will be used. As there are no real performance advantages.

The serialization of any other type is possible thanks to custom serializers.

Defining a schema

For a correct stringification of your json payload, a correct schema is mandatory. Defining a schema is pretty handy and not verbose at all.

1const { sjs } = require('slow-json-stringify');
2
3// schema definition
4const stringify = sjs({
5  a: attr('string'),
6  b: attr('number'),
7  c: attr('boolean'),
8});
9
10// then you can stringify anything with that structure.
11stringify({
12  a: 'world',
13  b: 42,
14  c: true,
15});
16
17// {"a":"world","b":42,"c":true}
18

Defining schema with simple array

When stringifying simple array JSON.stringify will be internally used.

1const { sjs } = require('slow-json-stringify');
2
3// schema definition
4const stringify = sjs({
5  a: attr('array'),
6});
7
8// then you can stringify anything with that structure.
9stringify({
10  a: [1, 2, 3, true, 'world'],
11});
12
13// {"a":[1,2,3,true,"world"]}
14

Defining schema with complex array

The attr helper accepts an additional sjs schema for array properties.

1const { sjs } = require('slow-json-stringify');
2
3// schema definition
4const stringify = sjs({
5  a: attr('array', sjs({
6    b: array('string'),
7    c: array('number'),
8  }))
9});
10
11// then you can stringify anything with that structure.
12stringify({
13  a: [{
14    b: 'ciao1',
15    c: 1,
16  }, {
17    b: 'ciao2',
18    c: 2,
19  }, {
20    b: 'ciao3',
21    c: 3,
22  }, {
23    b: 'ciao4',
24    c: 4, 
25  }],
26});
27
28// {"a":[{"b":"ciao1","c":1},{"b":"ciao2","c":2},{"b":"ciao3","c":3},{"b":"ciao4","c":4}]}
29

Defining schema with nested objects

Defining schemas with nested objects is pretty straightforward.

1const { sjs } = require('slow-json-stringify');
2
3// schema definition
4const stringify = sjs({
5  a: {
6    b: {
7      c: attr('string'),
8    },
9  },
10  d: {
11    e: attr('number'),
12  },
13});
14
15stringify({
16  a: {
17    b: {
18      c: 'hello',
19    },
20  },
21  d: {
22    e: 42,
23  },
24});
25
26// {"a":{"b":{"c":"hello"}},"d":{"e":42}}
27

Serializers

The attr helper accepts a serializer function. The serializer function gets invoked with the real value that should be stringified.

1serializer(rawValue)

Property serializers are useful to perform custom serialization on any provide type not natively supported even by JSON.stringify (Dates, regular expressions). They can be used also to skip property serializarion when returning undefined.

1const { sjs } = require('slow-json-stringify');
2
3// schema definition
4const stringify = sjs({
5  a: attr('number', (value) => {
6    if (value > 10) {
7      return value;
8    }
9
10    return undefined;
11  })
12});
13
14stringify({ a: 20 });
15// {"a":20}
16
17stringify({ a: 5 });
18// {}
19

use case: Serialization of any type.

1// DATES
2const stringify = sjs({
3  date: attr('string', (value) => value.toLocaleString()),
4});
5
6// REGEXP
7const stringify = sjs({
8  regexp: attr('string', (value) => value.toString()),
9});

use case: Customize payloads based on access rights.

1const stringify = sjs({
2  publicProp: attr('string'),
3  restrictedProp: attr('string', (value) => isAdmin ? value : undefined),
4});

use case: Value formatting

1const stringify = sjs({
2  prop: attr('string', (value) => value.toUpperCase()),
3});

API

sjs

param	type	required	default	spec
schema	object	yes	undefined	Schema that defines the stringification behavior.

attr

param	type	required	default	spec
type	string	yes	undefined	Type of the property.
serializer	function	no	identity	Function used for serializing / validating properties.

escape

param	type	required	default	spec
regex	Regular Expression	no	default regex	regex used to escape text

License

MIT.