SCALE Codecs for JavaScript and TypeScript

A TypeScript implementation of SCALE (Simple Concatenated Aggregate Little-Endian) transcoding (see Rust implementation here), which emphasizes JS-land representations and e2e type-safety. Faster than JSON, avsc, jsbin and protobuf (see benchmarks).

Setup

If you're using Deno, simply import via the deno.land/x specifier.

import * as $ from "https://deno.land/x/scale/mod.ts"

If you're using Node, install as follows.

npm install scale-codec

Then import as follows.

import * as $ from "scale-codec"

Usage

1. Import the library
2. Define a codec via the library's functions, whose names correspond to types
3. Utilize the codec you've defined

Example

import * as $ from "https://deno.land/x/scale/mod.ts"

const $superhero = $.object(
  $.field("pseudonym", $.str),
  $.optionalField("secretIdentity", $.str),
  $.field("superpowers", $.array($.str)),
)

const valueToEncode = {
  pseudonym: "Spider-Man",
  secretIdentity: "Peter Parker",
  superpowers: ["does whatever a spider can"],
}

const encodedBytes: Uint8Array = $superhero.encode(valueToEncode)
const decodedValue: Superhero = $superhero.decode(encodedBytes)

assertEquals(decodedValue, valueToEncode)

To extract the type from a given codec, you can use the Output utility type.

type Superhero = $.Output<typeof $superhero>
// {
//   pseudonym: string;
//   secretIdentity?: string | undefined;
//   superpowers: string[];
// }

You can also explicitly type the codec, which will validate that the inferred type aligns with the expected.

interface Superhero {
  pseudonym: string
  secretIdentity?: string
  superpowers: string[]
}

const $superhero: Codec<Superhero> = $.object(
  $.field("pseudonym", $.str),
  $.optionalField("secretIdentity", $.str),
  $.field("superpowers", $.array($.str)),
)

// @ts-expect-error
//   Type 'Codec<{ pseudonym: string; secretIdentity?: string | undefined; }>' is not assignable to type 'Codec<Superhero>'.
//     The types returned by 'decode(...)' are incompatible between these types.
//       Type '{ pseudonym: string; secretIdentity?: string | undefined; }' is not assignable to type 'Superhero'.
const $plebeianHero: Codec<Superhero> = $.object(
  $.field("pseudonym", $.str),
  $.optionalField("secretIdentity", $.str),
)

You can also validate a value against a codec using $.assert or $.is:

value // unknown
if ($.is($superhero, value)) {
  value // Superhero
}

value // unknown
$.assert($superhero, value)
value // Superhero

If $.assert fails, it will throw a ScaleAssertError detailing why the value was invalid.

Further examples can be found in the examples directory.

Codec Naming

This library adopts a convention of denoting codecs with a $ – $.foo for built-in codecs, and $foo for user-defined codecs. This makes codecs easily distinguishable from other values, and makes it easier to have codecs in scope with other variables:

interface Person { ... }
const $person = $.object(...)
const person = { ... }

Here, the type, codec, and a value can all coexist without clashing, without having to resort to wordy workarounds like personCodec.

The main other library this could possibly clash with is jQuery, and its usage has waned enough that this is not a serious problem.

While we recommend following this convention for consistency, you can, of course, adopt an alternative convention if the $ is problematic – $.foo can easily become s.foo or scale.foo with an alternate import name.

Asynchronous Encoding

Some codecs require asynchronous encoding. Calling .encode() on a codec will throw if it or another codec it calls is asynchronous. In this case, you must call .encodeAsync() instead, which returns a Promise<Uint8Array>. You can call .encodeAsync() on any codec; if it is a synchronous codec, it will simply resolve immediately.

Asynchronous decoding is not supported.

Custom Codecs

If your encoding/decoding logic is more complicated, you can create custom codecs with createCodec:

const $foo = $.createCodec<Foo>({
  _metadata: $.metadata("$foo"),

  // A static estimation of the encoded size, in bytes.
  // This can be either an under- or over- estimate.
  _staticSize: 123,
  _encode(buffer, value) {
    // Encode `value` into `buffer.array`, starting at `buffer.index`.
    // A `DataView` is also supplied as `buffer.view`.
    // At first, you may only write at most as many bytes as `_staticSize`.
    // After you write bytes, you must update `buffer.index` to be the first unwritten byte.

    // If you need to write more bytes, call `buffer.pushAlloc(size)`.
    // If you do this, you can then write at most `size` bytes,
    // and then you must call `buffer.popAlloc()`.

    // You can also call `buffer.insertArray()` to insert an array without consuming any bytes.

    // You can delegate to another codec by calling `$bar._encode(buffer, bar)`.
    // Before doing so, you must ensure that `$bar._staticSize` bytes are free,
    // either by including it in `_staticSize` or by calling `buffer.pushAlloc()`.
    // Note that you should use `_encode` and not `encode`.

    // See the `EncodeBuffer` class for information on other methods.

    // ...
  },

  _decode(buffer) {
    // Decode `value` from `buffer.array`, starting at `buffer.index`.
    // A `DataView` is also supplied as `buffer.view`.
    // After you read bytes, you must update `buffer.index` to be the first unread byte.

    // You can delegate to another codec by calling `$bar._decode(buffer)`.
    // Note that you should use `_decode` and not `decode`.

    // ...
    return value
  },

  _assert(assert) {
    // Validate that `assert.value` is valid for this codec.
    // `assert` exposes various utility methods, such as `assert.instanceof`.
    // See the `AssertState` class for information on other methods.

    // You can delegate to another codec by calling `$bar._assert(assert)` or `$bar._assert(assert.access("key"))`.
    // Any errors thrown should be an instance of `$.ScaleAssertError`, and should use `assert.path`.

    // ...
  },
})