PSON is a super efficient binary serialization format for JSON focused on minimal encoding size.
- Everything has been ported to TypeScript
- Integers between 2^31 and 2^53 are now encoded correctly
- Encoding is now 300% faster
- Re-introduce exclusion capability without affecting performance too much
- Optimize decoder
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
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 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 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.
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)
$ npm install @as-com/pson
const {ProgressivePair, ...} = require("@as-com/pson");
...
import {ProgressivePair} from "@as-com/pson";
// 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);
...
});
The API is pretty much straight forward:
PSON.Pair#encode(json: *): ByteBuffer
encodes JSON to PSON dataPSON.Pair#toBuffer(json: *): Buffer
encodes straight to a node.js BufferPSON.Pair#toArrayBuffer(json: *): ArrayBuffer
encodes straight to an ArrayBuffer
PSON.Pair#decode(pson: ByteBuffer|Buffer|ArrayBuffer): *
decodes PSON data to JSON
new ProgressivePair([initialDictionary: Array.<string>])
constructs a new progressive encoder and decoder pair with an automatically filling keyword dictionaryProgressivePair#exclude(obj: Object)
Excludes an object's and its children's keywords from being added to the progressive dictionaryProgressivePair#include(obj: Object)
Undoes the former
new StaticPair([dictionary: Array.<string>])
constructs a new static encoder and decoder pair with a static (or empty) dictionary
License: Apache License, Version 2.0