Skip to content

looker-open-source/chatty

Repository files navigation

chatty

A simple web browser iframe host/client channel message manager. It uses MessageChannels to avoid cross-talk between multiple iframes. It allows configuring the iframe to run in sandboxed mode.

Basic use

A user first initiates the creation of a client iframe using the createHost(url) method, adding event handlers using on(eventName, data). They then creates the iframe using build(), and opens a communication channel using connect(). Once the channel opens, the user can send messages to the client with send(eventName, data)

import { Chatty } from 'chatty'

Chatty.createHost('//example.com/client.html')
  .on(Actions.SET_STATUS, (msg: Msg) => {
    const status: Element = document.querySelector('#host-status')!
    status.innerHTML = `${msg.status} 1`
  })
  .build()
  .connect()
  .then((client) => {
    document.querySelector('#change-status')!.addEventListener('click', () => {
      client.send(Actions.SET_STATUS, { status: 'Message to client 1' })
    })
  })
  .catch(console.error)

The client iframe can also be created using source from the createHostFromSource(source) method.

import { Chatty } from 'chatty'

Chatty.createHostFromSource(`
      <html>
        <body>
          <script src='//example.com/client.js' type="application/javascript" />
        </body>
      </html>
  `)

The client iframe creates its client using createClient(). It also adds event listeners, builds the client and connects. Once connected, it can send messages to its host.

import { Chatty } from 'chatty'

Chatty.createClient()
  .on(Actions.SET_STATUS, (msg: Msg) => {
    const status = document.querySelector('#client-status')!
    status.innerHTML = msg.status
  })
  .build()
  .connect()
  .then((host) => {
    document.querySelector('#change-status')!.addEventListener('click', () => {
      host.send(Actions.SET_STATUS, { status: 'click from client' })
    })
  })
  .catch(console.error)

Sending and receiving

Both the host and the client can send a message and wait for a response. The sendAndReceive() method returns a promise that is resolved with a values returned by the event listeners on the client or host.

For example, a host can request that the client return its title.

import { Chatty } from 'chatty'

Chatty.createHost('//example.com/client.html')
  .build()
  .connect()
  .then((client) => {
    document.querySelector('#get-title')!.addEventListener('click', () => {
      client.sendAndReceive(Actions.GET_TITLE).then((payload: any[]) => {
        const title: Element = document.querySelector('#got-title')!
        title.innerHTML = payload[0]
      })
    })
  })
  .catch(console.error)

The client simply returns the text value of its title in the event handler.

import { Chatty } from 'chatty'

Chatty.createClient()
  .on(Actions.GET_TITLE, () => {
    return document.querySelector('title')!.text
  })
  .build()
  .connect()
  .catch(console.error)

The results provided by the promise are an array because their may be multiple handlers for a given event. If there are no event handlers for a given action the array will be empty.

Sending and receiving asynchronous responses

The sendAndReceive method can also be used for data that needs to be retrieved asynchronously. In this scenario the target function must return a Promise.

In the following example, the host requests that the client return some data that is to be retrieved asynchronously.

import { Chatty } from 'chatty'

Chatty.createHost('//example.com/client.html')
  .build()
  .connect()
  .then((client) => {
    document.querySelector('#get-title')!.addEventListener('click', () => {
      client.sendAndReceive(Actions.GET_TITLE).then((payload: any[]) => {
        const title: Element = document.querySelector('#got-title')!
        title.innerHTML = payload[0]
      })
    })
  })
  .catch(console.error)

The client message handler returns a Promise.

import { Chatty } from 'chatty'

Chatty.createClient()
  .on(Actions.GET_TITLE, () => {
    return new Promise((resolve, reject) => {
      setTimeout(() => {
        resolve(document.querySelector('title')!.text)
      }, 200)
    })
  })
  .build()
  .connect()
  .catch(console.error)

Sending and receiving timeouts

By default, sendAndReceive will timeout and throw an Error if a response is not received within the time specifed by the builder withDefaultTimeout method which defaults to 30 seconds. sendAndReceive will NOT timeout if withDefaultTimeout is set to a negative number. Alternatively an AbortSignal can be added to an Options object passed as the last argument of the sendAndReceive call. In this scenario the default timeout is ignored. The caller may then use the AbortController to terminate the sendAndReceive call.

In following example, the error message will be displayed in the title because the timeout that fires the abort will trigger before the timeout that returns the title.

import { Chatty } from 'chatty'

const abortController = new AbortController()
setTimeout(() => {
  abortController.abort()
}, 100)
Chatty.createHost('//example.com/client.html')
  .build()
  .connect()
  .then((client) => {
    document.querySelector('#get-title')!.addEventListener('click', () => {
      client
        .sendAndReceive(Actions.GET_TITLE, { signal: abortController.signal })
        .then((payload: any[]) => {
          const title: Element = document.querySelector('#got-title')!
          title.innerHTML = payload[0]
        })
        .catch((error: Error) => {
          const title: Element = document.querySelector('#got-title')!
          title.innerHTML = error.message
        })
    })
  })
  .catch(console.error)

The client message handler returns a Promise but its response will be ignored as the request will be aborted before the timer triggers.

import { Chatty } from 'chatty'

Chatty.createClient()
  .on(Actions.GET_TITLE, () => {
    return new Promise((resolve, reject) => {
      setTimeout(() => {
        resolve(document.querySelector('title')!.text)
      }, 200)
    })
  })
  .build()
  .connect()
  .catch(console.error)

Sending and receiving abort propagation

By default, if a signal is provided to sendAndReceive and it is aborted, the signal is NOT propagated to the message receiver. This behavior can be changed by setting propagateSignal to true in the Options object. When set, the target handler will receive an AbortSignal as the last argument of the handler.

This example demonstrates the use of propagateSignal.

const abortController = new AbortController()
setTimeout(() => {
  abortController.abort('100ms timeout')
}, 100)
client
  .sendAndReceive(
    Actions.PROPAGATED_ABORT_SIGNAL,
    { status: 'This message should not be displayed' },
    {
      signal: abortController.signal,
      propagateSignal: true,
    }
  )
  .then((payload: any[]) => {
    document.querySelector(`#got-propagate-abort-${id}`)!.innerHTML = payload[0]
  })
  .catch((error) => {
    document.querySelector(`#got-propagate-abort-${id}`)!.innerHTML =
      'error occured - see console'
    console.error('error occured', error)
  })

Notice how the message receiver clears the timer if an AbortSignal is received.

Chatty.createClient()
  .on(Actions.PROPAGATED_ABORT_SIGNAL, (msg: Msg, signal: AbortSignal) => {
    return new Promise((resolve, reject) => {
      const timeoutId = setTimeout(() => {
        const status = document.querySelector('#client-status')!
        status.innerHTML = msg.status
        resolve(status.innerHTML)
      }, 200)
      signal.addEventListener('abort', (event) => {
        clearTimeout(timeoutId)
        const status = document.querySelector('#client-status')!
        status.innerHTML = `Request aborted ${
          (event.target as AbortSignal).reason
        }`
      })
    })
  })
  .build()
  .connect()
  .catch(console.error)

Getting Started

  1. Make sure you have node and npm versions installed per package.json's "engines" field.
  2. npm install
  3. npm test
  4. npm start
  5. Happy hacking!

Repository Layout

  • /src - This is where you should do all the work on Chatty.
  • /lib - This is the built output generated by running npm run build. No editing should be done here.
  • /demo - This is what is hosted by WebpackDevServer via npm start. Use this to build a demo and test Chatty in real time (no need to refresh the page manually or restart the dev server, it does that for you).

NPM Commands

  • npm run build - runs the Typescript compiler, outputting all generated source files to /lib. Run this when creating a new build to distribute on github.
  • npm run lint - runs the ts linter
  • npm run lint-fix - runs the ts linter and attempts to auto fix problems
  • npm start - starts a dev server mounted on /demo.
  • npm test - runs the test suite for Chatty.