@benfen/bcs v0.1.0

BCS - Binary Canonical Serialization

This small and lightweight library implements Binary Canonical Serialization (BCS) in JavaScript, making BCS available in both Browser and NodeJS environments.

Install

To install, add the @benfen/bcs package to your project:

npm i @benfen/bcs

Quickstart

import { BCS, getSuiMoveConfig } from '@benfen/bcs';

const bcs = new BCS(getSuiMoveConfig());

// registering types so we can use them
bcs.registerAlias('UID', BCS.ADDRESS);
bcs.registerEnumType('Option<T>', {
	none: null,
	some: 'T',
});
bcs.registerStructType('Coin', {
	id: 'UID',
	value: BCS.U64,
});

// deserialization: BCS bytes into Coin
let bcsBytes = bcs
	.ser('Coin', {
		id: '0000000000000000000000000000000000000000000000000000000000000001',
		value: 1000000n,
	})
	.toBytes();
let coin = bcs.de('Coin', bcsBytes, 'hex');

// serialization: Object into bytes - an Option with <T = Coin>
let data = bcs.ser('Option<Coin>', { some: coin }).toString('hex');

console.log(data);

Description

BCS defines the way the data is serialized, and the result contains no type information. To be able to serialize the data and later deserialize, a schema has to be created (based on the built-in primitives, such as address or u64). There's no tip in the serialized bytes on what they mean, so the receiving part needs to know how to treat it.

Configuration

BCS constructor is configurable for the target. The following parameters are available for custom configuration:

// Example: All options used
import { BCS } from '@benfen/bcs';

const SUI_ADDRESS_LENGTH = 32;
const bcs = new BCS({
	vectorType: 'vector<T>',
	addressLength: SUI_ADDRESS_LENGTH,
	addressEncoding: 'hex',
	genericSeparators: ['<', '>'],
	types: {
		// define schema in the initializer
		structs: {
			User: {
				name: BCS.STRING,
				age: BCS.U8,
			},
		},
		enums: {},
		aliases: { hex: BCS.HEX },
	},
	withPrimitives: true,
});

let bytes = bcs.ser('User', { name: 'Adam', age: '30' }).toString('base64');

console.log(bytes);

For Sui Move there's already a pre-built configuration which can be used through the getSuiMoveConfig() call.

// Example: Sui Move Config
import { BCS, getSuiMoveConfig } from '@benfen/bcs';

const bcs = new BCS(getSuiMoveConfig());

// use bcs.ser() to serialize data
const val = [1, 2, 3, 4];
const ser = bcs.ser(['vector', BCS.U8], val).toBytes();

// use bcs.de() to deserialize data
const res = bcs.de(['vector', BCS.U8], ser);

console.assert(res.toString() === val.toString());

Similar configuration exists for Rust, the difference is the Vec<T> for vectors and address (being a special Move type) is not needed:

// Example: Rust Config
import { BCS, getRustConfig } from '@benfen/bcs';

const bcs = new BCS(getRustConfig());
const val = [1, 2, 3, 4];
const ser = bcs.ser(['Vec', BCS.U8], val).toBytes();
const res = bcs.de(['Vec', BCS.U8], ser);

console.assert(res.toString() === val.toString());

Built-in types

By default, BCS will have a set of built-in type definitions and handy abstractions; all of them are supported in Move.

Supported integer types are: u8, u16, u32, u64, u128 and u256. Constants BCS.U8 to BCS.U256 are provided by the library.

All of the type usage examples below can be used for bcs.de(<type>, ...) as well.

// Example: Primitive types
import { BCS, getSuiMoveConfig } from '@benfen/bcs';
const bcs = new BCS(getSuiMoveConfig());

// Integers
let _u8 = bcs.ser(BCS.U8, 100).toBytes();
let _u64 = bcs.ser(BCS.U64, 1000000n).toString('hex');
let _u128 = bcs.ser(BCS.U128, '100000010000001000000').toString('base64');

// Other types
let _bool = bcs.ser(BCS.BOOL, true).toString('hex');
let _addr = bcs.ser(BCS.ADDRESS, '0000000000000000000000000000000000000001').toBytes();
let _str = bcs.ser(BCS.STRING, 'this is an ascii string').toBytes();

// Vectors (vector<T>)
let _u8_vec = bcs.ser(['vector', BCS.U8], [1, 2, 3, 4, 5, 6, 7]).toBytes();
let _bool_vec = bcs.ser(['vector', BCS.BOOL], [true, true, false]).toBytes();
let _str_vec = bcs.ser('vector<bool>', ['string1', 'string2', 'string3']).toBytes();

// Even vector of vector (...of vector) is an option
let _matrix = bcs
	.ser('vector<vector<u8>>', [
		[0, 0, 0],
		[1, 1, 1],
		[2, 2, 2],
	])
	.toBytes();

Ser/de and formatting

To serialize and deserialize data to and from BCS there are two methods: bcs.ser() and bcs.de().

// Example: Ser/de and Encoding
import { BCS, getSuiMoveConfig, BcsWriter } from '@benfen/bcs';
const bcs = new BCS(getSuiMoveConfig());

// bcs.ser() returns an instance of BcsWriter which can be converted to bytes or a string
let bcsWriter: BcsWriter = bcs.ser(BCS.STRING, 'this is a string');

// writer.toBytes() returns a Uint8Array
let bytes: Uint8Array = bcsWriter.toBytes();

// custom encodings can be chosen when needed (just like Buffer)
let hex: string = bcsWriter.toString('hex');
let base64: string = bcsWriter.toString('base64');
let base58: string = bcsWriter.toString('base58');

// bcs.de() reads BCS data and returns the value
// by default it expects data to be `Uint8Array`
let str1 = bcs.de(BCS.STRING, bytes);

// alternatively, an encoding of input can be specified
let str2 = bcs.de(BCS.STRING, hex, 'hex');
let str3 = bcs.de(BCS.STRING, base64, 'base64');
let str4 = bcs.de(BCS.STRING, base58, 'base58');

console.assert((str1 == str2) == (str3 == str4), 'Result is the same');

Registering new types

Tip: all registering methods start with bcs.register* (eg bcs.registerStructType).

Alias

Alias is a way to create custom name for a registered type. It is helpful for fine-tuning a predefined schema without making changes deep in the tree.

// Example: Alias
import { BCS, getSuiMoveConfig } from '@benfen/bcs';
const bcs = new BCS(getSuiMoveConfig());

bcs.registerAlias('ObjectDigest', BCS.BASE58);

// ObjectDigest is now treated as base58 string
let _b58 = bcs.ser('ObjectDigest', 'Ldp').toBytes();

// we can override already existing definition to make it a HEX string
bcs.registerAlias('ObjectDigest', BCS.HEX);

let _hex = bcs.ser('ObjectDigest', 'C0FFEE').toBytes();

Struct

Structs are the most common way of working with data; in BCS, a struct is simply a sequence of base types.

// Example: Struct
import { BCS, getSuiMoveConfig } from '@benfen/bcs';
const bcs = new BCS(getSuiMoveConfig());

// register a custom type (it becomes available for using)
bcs.registerStructType('Balance', {
	value: BCS.U64,
});

bcs.registerStructType('Coin', {
	id: BCS.ADDRESS,
	// reference another registered type
	balance: 'Balance',
});

// value passed into ser function has to have the same
// structure as the definition
let _bytes = bcs
	.ser('Coin', {
		id: '0x0000000000000000000000000000000000000000000000000000000000000005',
		balance: {
			value: 100000000n,
		},
	})
	.toBytes();

Using Generics

To define a generic struct or an enum, pass the type parameters. It can either be done as a part of a string or as an Array. See below:

// Example: Generics
import { BCS, getSuiMoveConfig } from '@benfen/bcs';
const bcs = new BCS(getSuiMoveConfig());

// Container -> the name of the type
// T -> type parameter which has to be passed in `ser()` or `de()` methods
// If you're not familiar with generics, treat them as type Templates
bcs.registerStructType(['Container', 'T'], {
	contents: 'T',
});

// When serializing, we have to pass the type to use for `T`
bcs
	.ser(['Container', BCS.U8], {
		contents: 100,
	})
	.toString('hex');

// Reusing the same Container type with different contents.
// Mind that generics need to be passed as Array after the main type.
bcs
	.ser(['Container', ['vector', BCS.BOOL]], {
		contents: [true, false, true],
	})
	.toString('hex');

// Using multiple generics - you can use any string for convenience and
// readability. See how we also use array notation for a field definition.
bcs.registerStructType(['VecMap', 'Key', 'Val'], {
	keys: ['vector', 'Key'],
	values: ['vector', 'Val'],
});

// To serialize VecMap, we can use:
bcs.ser(['VecMap', BCS.STRING, BCS.STRING], {
	keys: ['key1', 'key2', 'key3'],
	values: ['value1', 'value2', 'value3'],
});

Enum

In BCS enums are encoded in a special way - first byte marks the order and then the value. Enum is an object, only one property of which is used; if an invariant is empty, null should be used to mark it (see Option<T> below).

// Example: Enum
import { BCS, getSuiMoveConfig } from '@benfen/bcs';
const bcs = new BCS(getSuiMoveConfig());

bcs.registerEnumType('Option<T>', {
	none: null,
	some: 'T',
});

bcs.registerEnumType('TransactionType', {
	single: 'vector<u8>',
	batch: 'vector<vector<u8>>',
});

// any truthy value marks empty in struct value
let _optionNone = bcs.ser('Option<TransactionType>', {
	none: true,
});

// some now contains a value of type TransactionType
let _optionTx = bcs.ser('Option<TransactionType>', {
	some: {
		single: [1, 2, 3, 4, 5, 6],
	},
});

// same type signature but a different enum invariant - batch
let _optionTxBatch = bcs.ser('Option<TransactionType>', {
	some: {
		batch: [
			[1, 2, 3, 4, 5, 6],
			[1, 2, 3, 4, 5, 6],
		],
	},
});

Inline (de)serialization

Sometimes it is useful to get a value without registering a new struct. For that inline struct definition can be used.

Nested struct definitions are not yet supported, only first level properties can be used (but they can reference any type, including other struct types).

// Example: Inline Struct
import { BCS, getSuiMoveConfig } from '@benfen/bcs';
const bcs = new BCS(getSuiMoveConfig());

// Some value we want to serialize
const coin = {
	id: '0000000000000000000000000000000000000000000000000000000000000005',
	value: 1111333333222n,
};

// Instead of defining a type we pass struct schema as the first argument
let coin_bytes = bcs.ser({ id: BCS.ADDRESS, value: BCS.U64 }, coin).toBytes();

// Same with deserialization
let coin_restored = bcs.de({ id: BCS.ADDRESS, value: BCS.U64 }, coin_bytes);

console.assert(coin.id == coin_restored.id, '`id` must match');
console.assert(coin.value == coin_restored.value, '`value` must match');

Aligning schema with Move

Currently, main applications of this library are:

1. Serializing transactions and data passed into a transaction
2. Deserializing onchain data for performance and formatting reasons
3. Deserializing events

In this library, all of the primitive Move types are present as built-ins, however, there's a set of special types in Sui which can be simplified to a primitive.

// Definition in Move which we want to read in JS
module me::example {
    struct Metadata has store {
        name: std::ascii::String,
    }

    struct ChainObject has key {
        id: sui::object::UID,
        owner: address,
        meta: Metadata
    }
    // ...
}

Definition for the above should be the following:

// Example: Simplifying UID
import { BCS, getSuiMoveConfig } from '@benfen/bcs';
const bcs = new BCS(getSuiMoveConfig());

// If there's a deep nested struct we can ignore Move type
// structure and use only the value.
bcs.registerAlias('UID', BCS.ADDRESS);

// Simply follow the definition onchain
bcs.registerStructType('Metadata', {
	name: BCS.STRING,
});

// Same for the main object that we intend to read
bcs.registerStructType('ChainObject', {
	id: 'UID',
	owner: BCS.ADDRESS,
	meta: 'Metadata',
});

struct ID has store, copy, drop { bytes: address }

// { id: { bytes: '0x.....' } }