AppLife

♻️ Handle the lifecyle of your application

Getting started

import AL from "applife"

const app = new AL({
  // first load your environment
  config: { up: loadenv },

  // when config is done, open pg connection
  pg: {
    needs: ["config"], // wait for config to have run
    up: async ({ config }) => { // each steps receive the full context
      const client = new Client();
      await client.connect();
      return client
    },
    down: ({ pg }) => pg.close(), // here pg is what was instanciated in `up`
    after: "start" // will only close pg AFTER the up stage has been downed
  },

  http: {
    needs: ["config"],
    up: ({ config }) => new Koa().use(/* ... */),
  },

  socket: {
    needs: ["http"],
    up: ({ http, pg }) => new Server(require("http").createServer(http.callback()))
  },

  start: {
    needs: ["pg", "http", "socket"],
    up: ({ http, config }) => http.listen(port),
    down: ({ start }) => up.close()
  }
})

app.on("stopped", signal => console.log(`App stopped because ${signal} was received`))
app.start()
  .then(({ config }) => console.log(`App started on port ${config.port}`))
  .catch(e => console.error("Could not start app", e))

This will start the config stage, then the pg and http stages, then the socket stage and finally if the start stage.

If your app receive SIGINT, SIGTERM, unhandledRejection or uncaughtException, (or if you call app.close()), start will be downed first then pg.

API

AppLife#constructor<T extends {[key: string]: unknown}>(dependencies)

Setup the the applife and load dependencies

• T extends {[key: string]: unknown} is the object that represent your state, once fully loaded.

  T can be infered from usage. For example if you have:

  const app = new Applife({
    a: { up: () => Promise.resolve("A") },
    b: { up: () => Promise.resolve(42) },
  })
  // T: { a: string, b: number }
  

• dependencies: {[key: keyof T]: dependency}, an object that represent the list of dependencies. The key is important as each dependency will receive Pick<T, LoadedEntries>

  • dependency is an object with the following keys

    • needs?: keyof T | Array<keyof T>: an optional array of steps which must resolve BEFORE calling the up method

    • up?: (loaded: Pick<T, LoadedEntries>) => Promise<T[step]>: the function to boot the current stage (once all in needs resolved).
      This function takes all loaded properties and MUST return a promise.
      up is optional because you may only want to specify a shutdown step.

    • after?: keyof T | Array<keyof T>: an optional array of steps which must resolve BEFORE calling the down method

    • down?: (loaded: Pick<T, LoadedEntries>) => Promise<T[step]>: The function to call to gracefully shutdown this step.
      This function will only be called AFTER all steps defined in after.
      This MUST return a promise

AppLife.start() => Promise<T>

Starts your application (run all the up functions) and returns loaded values.

If any dependency fails to load, start will throw a Boot sequence failed with a list of all errors that were caught in Error.details: Error[]
The throw will occur AFTER all dependencies are successfuly unloaded

start will return your resolved dependencies

AppLife.stop() => Promise<void>

Stop your application (run all the down functions).

stop will only stop a dependency that was started.

AppLife.run() => Promise<void>

Treat your application as if it were a one of application.

Basically is app.start().then(() => app.stop())

Handles dependencies error the same way start would.

emit stopped(reason)

After the app stops, it emits the stopped event with one of the following reasons:

• SIGTERM received the SIGTERM signal
• SIGINT received the SIGINT signal
• uncaughtException an exception was not handled by your code
• unhandledRejection a promise rejection was not handled by your code
• stop - if the stop method was called