Awesome

PSON is a super efficient binary serialization format for JSON focused on minimal encoding size.

How does it work?

PSON combines the best of JSON, BJSON, ProtoBuf and a bit of ZIP to achieve a superior small footprint on the network level. Basic constants and small integer values are efficiently encoded as a single byte. Other integer values are always encoded as variable length integers. Additionally it comes with progressive and static dictionaries to reduce data redundancy to a minimum. In a nutshell:

246 single byte values
Base 128 variable length integers (varints) as in protobuf
32 bit floats instead of 64 bit doubles if possible without information loss
Progressive and static dictionaries
Raw binary data support
Long support

Reference implementation

This repository contains a plain node.js/CommonJS, RequireJS/AMD and Browser compatible JavaScript implementation of the PSON specification on top of ByteBuffer.js and optionally Long.js:

A PSON.StaticPair contains the PSON encoder and decoder for a static (or empty) dictionary and can be shared between all connections. It's recommended for production.

A PSON.ProgressivePair contains the PSON encoder and decoder for a progressive (automatically filling) dictionary. On the one hand this requires no dictionary work from the developer but on the other requires one pair per connection.

tl;dr Numbers, please!

The test suite contains the following basic example message:

{
    "hello": "world!",
    "time": 1234567890,
    "float": 0.01234,
    "boolean": true,
    "otherbool": false,
    "null": null,
    "obj": {
        "what": "that"
    },
    "arr": [1,2,3]
}

JSON stringify: 133 bytes
PSON without a dictionary: 103 bytes (about 22% smaller than JSON)
PSON with a progressive dictionary: 103 bytes for the first and 59 bytes for each subsequent message (about 22% smaller for the first and about 55% smaller for each subsequent message than JSON.
PSON with the same but static dictionary: 59 bytes for each message (about 55% smaller than JSON)

 F6 08 FE 00 FC 06 77 6F 72 6C 64 21 FE 01 F8 A4   ......world!....
 8B B0 99 79 FE 02 FB F6 0B 76 C3 B6 45 89 3F FE   ...y.....v..E.?.
 03 F1 FE 04 F2 FE 05 F0 FE 06 F6 01 FE 07 FC 04   ................
 74 68 61 74 FE 08 F7 03 02 04 06                  that.......

Another example that's also contained in the test suite is encoding our package.json, which is of course a string value centered file, to PSON using a general purpose static dictionary:

JSON stringify: 813 bytes
PSON with empty dict: 760 bytes (about 6% smaller than JSON)
PSON with static dict: 613 bytes (about 24% smaller than JSON)

Usage

node.js/CommonJS

npm install pson

var PSON = require("pson");
...

RequireJS/AMD

require.config({
    ...
    "paths": {
        "Long": "/path/to/Long.js", // optional
        "ByteBuffer": "/path/to/ByteBufferAB.js",
        "PSON": "/path/to/PSON.js"
    },
    ...
});
require(["PSON"], function(PSON) {
    ...
});

Browser

<script src="Long.min.js"></script>
<script src="ByteBufferAB.min.js"></script>
<script src="PSON.min.js"></script>

var PSON = dcodeIO.PSON;
...

Example

// Sender
var initialDictionary = ["hello"];
var pson = new PSON.ProgressivePair(initialDictionary);
var data = { "hello": "world!" };
var buffer = pson.encode(data);
someSocket.send(buffer);

// Receiver
var initialDictionary = ["hello"];
var pson = new PSON.ProgressivePair(initialDictionary);
someSocket.on("data", function(data) {
    data = pson.decode(data);
    ...
});

API

The API is pretty much straight forward:

PSON.Pair#encode(json: *): ByteBuffer encodes JSON to PSON data
- PSON.Pair#toBuffer(json: *): Buffer encodes straight to a node.js Buffer
- PSON.Pair#toArrayBuffer(json: *): ArrayBuffer encodes straight to an ArrayBuffer
PSON.Pair#decode(pson: ByteBuffer|Buffer|ArrayBuffer): * decodes PSON data to JSON

Progressive

new PSON.ProgressivePair([initialDictionary: Array.<string>]) constructs a new progressive encoder and decoder pair with an automatically filling keyword dictionary
PSON.ProgressivePair#exclude(obj: Object) Excludes an object's and its children's keywords from being added to the progressive dictionary
PSON.ProgressivePair#include(obj: Object) Undoes the former

Static

new PSON.StaticPair([dictionary: Array.<string>]) constructs a new static encoder and decoder pair with a static (or empty) dictionary

Downloads

Documentation

License: Apache License, Version 2.0