Skip to content

Latest commit

 

History

History
406 lines (305 loc) · 8.66 KB

README.md

File metadata and controls

406 lines (305 loc) · 8.66 KB

Borsh TS

Project license Project license NPM version Size on NPM

Borsh TS is a Typescript implementation of the Borsh binary serialization format for TypeScript projects. The motivation behind this library is to provide more convinient methods using field and class decorators.

Borsh stands for Binary Object Representation Serializer for Hashing. It is meant to be used in security-critical projects as it prioritizes consistency, safety, speed, and comes with a strict specification.

This implementation is performant, in fact, It slightly outperforms protobuf.js benchmark 2 (~20% faster), benchmark 3 (~15% faster)

How borsh-ts differs from borsh-js

  • Schema is defined using decorators rather than building a map. The schema is stored alongside the class behind the scenes so there is no longer need to pass it during serialization and deserialization.
  • Big number are interpreted with bigint rather than BN (bn.js)
  • No dependency on Buffer
  • ESM and CJS build
  • Stricter validation checks during serialization and deserialization

Installation

npm install @dao-xyz/borsh

or

yarn add @dao-xyz/borsh

Serializing and deserializing

Serializing an object

SomeClass class is decorated using decorators explained later

import {
  deserialize,
  serialize,
  field,
  variant,
  vec,
  option
} from "@dao-xyz/borsh";

class SomeClass 
{
    @field({type: 'u8'})
    x: number

    @field({type: 'u64'})
    y: bigint

    @field({type: 'string'})
    z: string

    @field({type: option(vec('u32'))})
    q?: number[]

    constructor(data: SomeClass)
    {
       Object.assign(this, data)
    }
}

...

const value = new SomeClass({ x: 255, y: 20n, z: 'abc', q: [1, 2, 3] });

// Serialize 
const serialized = serialize(value); 

// Deserialize
const deserialized = deserialize(serialized,SomeClass);

Examples of schema generation using decorators

For more examples, see the tests.

Enum, with 2 variants

abstract class Super {}

@variant(0)
class Enum0 extends Super {

    @field({ type: "u8" })
    public a: number;

    constructor(a: number) {
        super();
        this.a = a;
    }
}

@variant(1)
class Enum1 extends Super {

    @field({ type: "u8" })
    public b: number;

    constructor(b: number) {
        super();
        this.b = b;
    }
}

class TestStruct {

    @field({ type: Super })
    public enum: Super;

    constructor(value: Super) {
        this.enum = value;
    }
}

Variants can be 'number', 'number[]' (represents nested Rust Enums) or 'string' (not part of the Borsh specification). i.e.

@variant(0)
class ClazzA
...
@variant([0,1])
class ClazzB
...
@variant("clazz c")
class ClazzC

Nested Schema generation for structs

class InnerStruct {

    @field({ type: 'u32' })
    public b: number;

}

class TestStruct {

    @field({ type: InnerStruct })
    public a: InnerStruct;

}

Strings

With Borsh specification, string sizes will be encoded with 'u32'

class TestStruct {
  @field({ type: 'string' })
  public string: string; 
}

You can override this (i.e. not longer adhere to the specification)

class TestStruct {
  @field({ type: string('u8')}) // less memory for small strings
  public string: string; 
}

Arrays

Dynamically sized

class TestStruct {
  @field({ type: Uint8Array })
  public vec: Uint8Array; 
}
class TestStruct {
  @field({ type: vec('u8') })
  public vec: Uint8Array; // Uint8Array will be created if type is u8
}
class TestStruct {
  @field({ type: vec('u32') })
  public vec: number[];
}

Custom size encoding, not compatible with Borsh specification

class TestStruct {
  @field({ type: vec('u64', 'u8') }) // Size will be encoded with u8 instead of u32
  public vec: bigint[]; 
}

Fixed length

class TestStruct {
  @field({ type: fixedArray('u8', 3) }) // Fixed array of length 3
  public fixedLengthArray: Uint8Array;  // Uint8Array will be created if type is u8
}
class TestStruct {
  @field({ type: fixedArray('u64', 3) }) // Fixed array of length 3
  public fixedLengthArray: bigint[];
}

Option

class TestStruct {
  @field({ type: option('u8') })
  public a: number | undefined;
}

Custom serialization and deserialization Override how one field is handled

class TestStruct {

    // Override ser/der of the number
    @field({
        serialize: (value: number, writer) => {
            writer.u16(value);
        },
        deserialize: (reader): number => {
            return reader.u16();
        },
    })
    public number: number;
    constructor(number: number) {
        this.number = number;
    }
}

const serialized = serialize(new TestStruct(3));
const deserialied = deserialize(serialized, TestStruct);
expect(deserialied.number).toEqual(3);

Override how one class is serialized

import { serializer } from '@dao-xyz/borsh'
class TestStruct {

  @field({type: 'u8'})
  public number: number;

  constructor(number: number) {
    this.number = number;
  }

  cache: Uint8Array | undefined;

  @serializer()
  override(writer: BinaryWriter, serialize: (obj: this) => Uint8Array) {
    if (!this.cache) {
      this.cache = serialize(this)
    }
    writer.set(this.cache)
  }
}

const obj = new TestStruct(3);
const serialized = serialize(obj);
const deserialied = deserialize(serialized, TestStruct);
expect(deserialied.number).toEqual(3);
expect(obj.cache).toBeDefined()

Inheritance

Schema generation is supported if deserialization is deterministic. In other words, all classes extending some super class needs to use discriminators/variants of the same type.

Example:

class A {
    @field({type: 'u8'})
    a: number 
}

@variant(0)
class B1 extends A{
    @field({type: 'u16'})
    b1: number 
}

@variant(1)
class B2 extends A{
    @field({type: 'u32'})
    b2: number 
}

Discriminator

It is possible to resolve the discriminator without serializing a class completely

import { getDiscriminator} from '@dao-xyz/borsh'

@variant([1, 2])
class A { }
class B extends A { }

@variant(3)
class C extends B { }

const discriminator = getDiscriminator(C);
expect(discriminator).toEqual(new Uint8Array([1, 2, 3]));

Explicit serialization order of fields

class TestStruct {
    @field({ type: 'u8', index: 1 })
    public a: number;


    @field({ type: 'u8', index: 0 })
    public b: number;
}

This will make b serialized into the buffer before a.

Validation

You can validate that classes have been decorated correctly:

validate([TestStruct])

Type Mappings

Borsh TypeScript
u8 integer number
u16 integer number
u32 integer number
u64 integer bigint
u128 integer bigint
u256 integer bigint
u512 integer bigint
f32 float number
f64 float number
byte arrays Uint8Array
UTF-8 string string
option undefined or type
map N/A
set N/A
structs any

Contributing

Install dependencies:

yarn install

Run tests:

yarn test

Run linter

yarn pretty

Benchmarks

See benchmark script here

  • json x 2,038,236 ops/sec ±0.13% (244 runs sampled)
  • borsh x 3,963,769 ops/sec ±0.38% (396 runs sampled)
  • protobujs x 3,221,585 ops/sec ±0.31% (394 runs sampled)

License

This repository is distributed under the terms of both the MIT license and the Apache License (Version 2.0). See LICENSE-MIT and LICENSE-APACHE for details.

For official releases see: [Borsh]: https://borsh.io