Skip to main content

Async Call

async-call-rpc is a JSON RPC server and client written in TypeScript for any ES6+ environment.

Code coverage GitHub Workflow Status npm ES2015+

CHANGELOG.md | Document of AsyncCall | Document of AsyncGeneratorCall | Playground

Chapters:

Features

  • Zero dependencies!
  • Running in any ES6+ environment (+globalThis), no requirement on any Web or Node API
  • Simple to define a server and simple to use as a client
  • Full TypeScript support
  • Support custom serializer to pass complex data types
  • Support async generator (Require both server and client supports 4 JSON RPC internal methods, and Symbol.asyncIterator, (async function* () {}).constructor.prototype available)

Cautions

  • NOT support ECMAScript 5 (ES6 Proxy is the core of this library)
  • This package is shipping ECMAScript 2018 syntax (including async function).
  • The async generator mode might leak memory on the server. Use it by your caution.
  • NOT support JSON RPC 1.0

The first concept: channel

The channel is the only thing you need to learn to use this library.

This library is designed to not rely on any specific platform. Only require things defined in the ECMAScript specification. In the ES spec, there is no I/O related API so it’s impossible to communicate with the outer world.

You need to implement one of the following interfaces:

There are some built-in channel you can simplify the usage.

The following document will assume you have defined your channel.

Example

Server example

// server.ts
export function add(x: number, y: number) {
    return x + y
}
export const sleep = (ms: number) => new Promise((resolve) => setTimeout(resolve, ms))

// init.ts
import { AsyncCall } from 'async-call-rpc'
import * as server from './server'
// create a server
AsyncCall(server, { channel })

Client example

import { AsyncCall } from 'async-call-rpc'
const server = AsyncCall<typeof server>({}, { channel })
server.add(2, 40).then(console.log) // 42

Isomorphic API

You can notice from the above example, define a server is using AsyncCall(serverImplementation, opt), define a client is using AsyncCall<typeof serverImplementation>({}, opt). So it is possible to define a server and a client at the same time.

Notifications and Batch requests

AsyncCall can send Notifications.

Using notifications means results or remote errors will be dropped. Local errors won’t be omitted, e.g. serializer error or network error.

import { AsyncCall, notify } from 'async-call-rpc'
const server = notify(AsyncCall<typeof server>({}, { channel }))
server.online().then(console.log) // undefined

AsyncCall can send batch request too.

import { AsyncCall, batch } from 'async-call-rpc'
const [server, emit, drop] = batch(AsyncCall<typeof server>({}, { channel }))
const a = server.req1() // pending
const b = server.req2() // pending
const c = server.req3() // pending
emit() // to send all pending requests
// request a, b, c sent

const d = server.req1() // pending
drop() // to drop all pending requests (and corresponding Promises)
// d rejected

Installation

Install through npm

npm i async-call-rpc

yarn add async-call-rpc

Import from browser or Deno

You can access https://www.jsdelivr.com/package/npm/async-call-rpc?path=out to get the latest URL and SRI.

(Supports type definition for deno out-of-box!)

import { AsyncCall } from 'https://cdn.jsdelivr.net/npm/async-call-rpc@latest/out/base.mjs'

UMD

<script src="https://cdn.jsdelivr.net/npm/async-call-rpc@2.0.1/out/base.js"></script>
<script>
    const { AsyncCall } = globalThis.AsyncCall
</script>

In other JS environment

Load the out/base.mjs (ES Module) or out/base.js (UMD, CommonJS or AMD) to your project.

Entries

This library has 2 entry. base and full. base is the default entry point. The full version includes the AsyncGeneratorCall but the base version doesn’t.

Browser / Deno

Please check out https://www.jsdelivr.com/package/npm/async-call-rpc?path=out

Node:

// Full version
require('async-rpc-call/full') // or
import * as RPC from 'async-rpc-call/full'

// Base version
require('async-rpc-call/base') // or
import * as RPC from 'async-rpc-call/base'

Builtin channels

They’re not part of the core library but provided as utils to increase usability.

WebSocket Server: Node

Entry point: async-call-rpc/utils/node/websocket.server.js.

Dependencies: ws

Example: ./examples/node.websocket.server.js

WebSocket Server: Deno

Entry point: https://cdn.jsdelivr.net/npm/async-call-rpc@3.2.0/utils/deno/websocket.server.ts

Example: ./examples/deno.websocket.server.ts

WebSocket Client: Web

Entry point: https://cdn.jsdelivr.net/npm/async-call-rpc@3.2.0/utils/web/websocket.client.js

Example: ./examples/browser.websocket.client.js

BroadcastChannel Server & Client: Web

Entry point: https://cdn.jsdelivr.net/npm/async-call-rpc@3.2.0/utils/web/broadcast.channel.js

Utils available if both server and client are created by this library

AsyncCall has some non-standard extensions to the JSON RPC specification that can help the library easier to use. Those features aren’t enabled by default.

Implemented JSON RPC internal methods

These four methods are used to implement AsyncGeneratorCall support.

interface JSONRPC_Internal_Methods {
    // These 4 methods represent the Async Iterator protocol in ECMAScript
    // this method starts an async iterator, return the id
    'rpc.async-iterator.start'(method: string, params: unknown[]): Promise<string>
    // this method executes `next` method on the previous iterator started by `rpc.async-iterator.start`
    'rpc.async-iterator.next'(id: string, value: unknown): Promise<IteratorResult<unknown>>
    // this method executes `return` method on the previous iterator started by `rpc.async-iterator.start`
    'rpc.async-iterator.return'(id: string, value: unknown): Promise<IteratorResult<unknown>>
    // this method executes `throw` method on the previous iterator started by `rpc.async-iterator.start`
    'rpc.async-iterator.throw'(id: string, value: unknown): Promise<IteratorResult<unknown>>
}

Non-standard extension to JSON RPC specification

remoteStack on Request object

This library can send the client the call stack to the server to make the logger better.

Controlled by option.log.sendLocalStack. Default to false.

interface JSONRPC_Request_object {
    // This property include the caller's stack.
    remoteStack?: string
}

“undef” on Response object

This is a non-standard property appears when using JSONSerialization due to JSON doesn’t support undefined. It’s a hint to the client, that the result is undefined.

This behavior is controlled by the 3rd parameter of JSONSerialization(replacerAndReceiver?, space?, undefinedKeepingBehavior?: false | “keep” | “null” = “null”). Default to "null". To turn on this feature to “keep” undefined values, change the 3rd option to “keep”.

interface JSONRPC_Response_object {
    // This property is a hint.
    // If the client is run in JavaScript, it should treat "result: null" as "result: undefined"
    undef?: boolean
}

The implementation-defined Error data

In the JSON RPC specification, this is implementation-defined. This is controlled by the option options.mapError

This library will try to “Recover” the Error object if there is enough information from another side.

interface JSONRPC_Error_object {
    // This property will help client to build a better Error object.
    data?: {
        stack?: string
        // Supported value for "type" field (Defined in ECMAScript specification):
        type?:
            | string
            | 'Error'
            | 'EvalError'
            | 'RangeError'
            | 'ReferenceError'
            | 'SyntaxError'
            | 'TypeError'
            | 'URIError'
            // Defined in HTML specification, only supported in Web
            | 'DOMException'
    }
}