V8 serialize is not deterministic

node version v10.17.0:
Linux 4.19.19-041919-lowlatency #201901310331 SMP PREEMPT Thu Jan 31 08:37:07 UTC 2019 x86_64 x86_64 x86_64 GNU/Linux:
npm version:

{ npm: '6.12.0',
  ares: '1.15.0',
  brotli: '1.0.7',
  cldr: '35.1',
  http_parser: '2.8.0',
  icu: '64.2',
  modules: '64',
  napi: '5',
  nghttp2: '1.39.2',
  node: '10.17.0',
  openssl: '1.1.1d',
  tz: '2019a',
  unicode: '12.1',
  uv: '1.28.0',
  v8: '6.8.275.32-node.54',
  zlib: '1.2.11' }

v8.serialize produce different output, still decodable but it's different bytes, and different length. it's matter for me.
same PC, same version.
just want to know: what cause that difference and how to avoid it.

const v8 = require('v8');

let data = {
    key: 'value',
    arrValue: [1,2,3,4],
    objValue: {
        nestedKey: 'value'
    }
};

const buffData = v8.serialize(data);
const dataCopy = v8.deserialize(buffData);

// it seems we still have same object
console.log(JSON.stringify(dataCopy) === JSON.stringify(data));

// but encoded buffer is different!
const buffEncoded = v8.serialize(Object.assign({}, dataCopy));
console.log(buffData.length !== buffEncoded.length);

console.log(buffData.toString('hex'));
console.log(buffEncoded.toString('hex'));

If you remove array element - it works like a charm again!
i mean

let data = {
    key: 'value',
    objValue: {
        nestedKey: 'value'
    }
};

encodes as expected

V8 uses different internal representations for some values, for example, Arrays in V8 can be dense or sparse, with special cases for integer arrays or floating-point-value arrays, and strings can have different internal representations as well, e.g. using UTF16 encoding or using ISO-8859-1.

Generally, you can’t rely on the exact output from v8.serialize(), that’s correct. If it’s important to you to have deterministic output, it might not be the right tool for you.

V8 uses different internal representations for some values, for example, Arrays in V8 can be dense or sparse, with special cases for integer arrays or floating-point-value arrays, and strings can have different internal representations as well, e.g. using UTF16 encoding or using ISO-8859-1.

it's ok, but i didn't expect diffrent output for "same" values.

doc: clarify that `v8.serialize()` is not deterministic

Navigation Menu

Repository navigation

V8 serialize is not deterministic #2448

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

V8 serialize is not deterministic #2448

Description

Activity

trueshura commented on Jan 28, 2020

addaleax commented on Jan 28, 2020

trueshura commented on Jan 28, 2020

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions