compress-json

compress-json

Store JSON data in space efficient manner.

Inspired by compressed-json and jsonpack.

This library is optimized to reduce represent json object in compact format, which can save network bandwidth and disk space. It is not optimized for writing nor querying throughput. Although the reduced IO may speed up usage of lmdb on frequently redundant data, that is not the design goal.

Features

• Object key order is preserved
• Repeated values are stored only once
• Numbers are encoded in base62 format (0-9A-Za-z)
• Support multiple storage backend
  • in-memory object / array / Map
  • localStorage
  • lmdb
  • leveldb (sync mode)
  • custom adapter

Installation

npm i -S compress-json

Usage

import { compress, decompress } from 'compress-json'

let data = { 
  user: 'Alice',
  // more fields of any json values (string, number, array, object, e.t.c.)
}

let compressed = compress(data) // the result is an array
fetch('/submit', {
  method: 'post',
  body: JSON.stringify(compressed) // convert into string if needed
})

let reversed = decompress(compressed)
data === reversed // will be false
JSON.stringify(data) === JSON.stringify(reversed) // will be true

Format

Sample data:

let longStr = 'A very very long string, that is repeated'
let data = {
  int: 42,
  float: 12.34,
  str: 'Alice',
  longStr,
  longNum: 9876543210.123455,
  bool: true,
  bool2: false,
  arr: [42, longStr],
  arr2: [42, longStr], // identical values will be deduplidated, including array and object
  obj: { // nested values are supported
    id: 123,
    name: 'Alice',
    role: [ 'Admin', 'User', 'Guest' ],
    longStr: 'A very very long string, that is repeated',
    longNum: 9876543210.123455
  },
  escape: [ 's|str', 'n|123', 'o|1', 'a|1', 'b|T', 'b|F' ]
}

Compressed data:

// [ encoded value array, root value index ]
let compressed = [
  [  // encoded value array
    'int', // string
    'float',
    'str',
    'longStr',
    'longNum',
    'bool',
    'bool2',
    'arr',
    'arr2',
    'obj',
    'escape',
    'a|0|1|2|3|4|5|6|7|8|9|A',
    'n|g', // number (integer) (base62-encoded)
    'n|C.h', // number (float) (integer part and decimals are base62-encoded separately)
    'Alice',
    'A very very long string, that is repeated',
    'n|AmOy42.2KCf',
    'b|T', // boolean (true)
    'b|F', // boolean (false)
    'a|C|F', // array
    'id',
    'name',
    'role',
    'a|K|L|M|3|4',
    'n|1z',
    'Admin',
    'User',
    'Guest',
    'a|P|Q|R',
    'o|N|O|E|S|F|G', // object
    's|s|str', // escaped string
    's|n|123', // escaped number
    's|o|1',
    's|a|1',
    's|b|T', // escaped boolean
    's|b|F',
    'a|U|V|W|X|Y|Z',
    'o|B|C|D|E|F|G|H|I|J|J|T|a'
  ],
  'b' // root value index
]

Benchmark

Test file: compress-test.ts

Sample data in use: json data of 109,164 threads and 724,905 post crawled from a discuz forum. Truncated in varies of size for testing.

Algorithms in comparison:

• JSON (JSON.stringify without indentation)
• compressed-json
• jsonpack
• compress-json (this library)

(Binary compression algorithm is not considered)

Compressed Size

sample	JSON	compressed-json	jsonpack	compress-json
all	263M	199M	-	176M
100,000	235M	178M	-	158M
50,000	70M	55M	-	50M
10,000	34M	26M	-	23M
2,000	6.6M	5.0M	5.3M	4.4M
1,000	4.8M	3.7M	3.8M	3.3M
100	335K	265K	271K	243K
10	4.0K	3.3K	3.0K	3.2K

Compression Time

sample	JSON	compressed-json	jsonpack	compress-json
all	1,654ms	12,674ms	timeout*	15,788ms
100,000	1,500ms	10,921ms	timeout*	12,715ms
50,000	462ms	3,047ms	timeout*	3,935ms
10,000	146ms	1,278ms	timeout*	1,733ms
2,000	35ms	328ms	21,018ms	456ms
1,000	20ms	270ms	12,960ms	390ms
100	1ms	18ms	47ms	37ms
10	0.3ms	1.8ms	1.6ms	1.9ms

*timeout: excess 1 minute

Decompress Time

sample	JSON	compressed-json	jsonpack	compress-json
all	1,908ms	4,611ms	-	9,225ms
100,000	1,744ms	3,740ms	-	7,576ms
50,000	558ms	1,066ms	-	2,452ms
10,000	173ms	460ms	-	1,055ms
2,000	47ms	108ms	189ms	317ms
1,000	34ms	90ms	160ms	263ms
100	2ms	11ms	16ms	19ms
10	2.9ms	2.1ms	1.1ms	1.3ms

License

BSD 2-Clause License (Free Open Sourced Software)