CODEWITHSUNDEEP
jsonmsgpack
Shankhoneer Chakrovarty
Shankhoneer Chakrovarty

Reputation: 685

Why msgpack is slower than json.parse?

I am investigating different data format (string vs masgpack byte) to encode and decode json efficiently. The jsons which I will recieve wont have any schema, atleast to keep it generic, I am assuming the jsons dont follow any specific schema.

Here is a small JS I used to evaluate:

// skip-eslint
var msgpack = require('msgpack5')() // namespace our extensions
  , encode  = msgpack.encode
  , decode  = msgpack.decode;

var helloWorldObj = { 'hello': 'world' };
var json = [];

var MAX_DECODE_REPEAT = 10000;
var MAX_REPEAT  = 100;

console.log('json,data,time');
//msgpack
for ( var i=0;i<MAX_REPEAT;i++) {
  json = [];
  // create i+1 length json array
  for(var k=0;k<=i;k++){
    json.push(helloWorldObj);
  }

  // encode the json array
  var encodedJson = encode(json);
  var start = new Date().getTime();
  // start decoding =>msgpack
  for(var count=0;count<MAX_DECODE_REPEAT;count++){
    decode(encodedJson);
  }
  var end = new Date().getTime();
  var time = end-start;
  console.log(json.length +',' + encodedJson.length + ','+time);

}

// JSON.parse
for ( var i=0;i<MAX_REPEAT;i++) {
  json = [];
  // create i+1 length json array
  for(var k=0;k<=i;k++){
    json.push(helloWorldObj);
  }

  // stringify the json array
  var jsonString = JSON.stringify(json);
  var start = new Date().getTime();
  // start decoding =>msgpack
  for(var count=0;count<MAX_DECODE_REPEAT;count++){
    JSON.parse(jsonString);
  }
  end = new Date().getTime();
  time = end-start;
  console.log(json.length +',' + jsonString.length*2 + ','+time);
}

The logic is simple: I am increasing the length of the array and trying to find out the time taken by the decoder to decode the array 10000 times.

To my surprise, msgpack is taking 10 times more than json.parse. To decode 100 length array msgpack takes 4694 ms whereas json.parse takes only 378 ms, even though msgpack compresses the array to 1303 bytes from 3602 bytes (assuming each character takes 2 bytes).

What I initially assumed is a smaller size to decode would mean smaller time to decode but its definitely not the case. Any idea why? Do you see any issue with the program to evaluate msgpack?

Thanks

Upvotes: 0

Views: 2266

Answers (1)

Wildfire
Wildfire

Reputation: 6418

You're observing this difference because JSON.parse uses optimized native implementation under the hood, whereas msgpack5 is a pure-JS library. In other languages (i.e., C, Java, etc) MsgPack will be slightly faster in most situations (for example, see this benchmark: https://github.com/eishay/jvm-serializers/wiki).

There are few corner cases when there'll be significant difference. For example, it's likely that msgpack will be faster when serializing numbers/booleans.

Upvotes: 1

Related Questions