Huge News!Announcing our $40M Series B led by Abstract Ventures.Learn More
Socket
Sign inDemoInstall
Socket

iti

Package Overview
Dependencies
Maintainers
1
Versions
15
Alerts
File Explorer

Advanced tools

Socket logo

Install Socket

Detect and block malicious and high-risk dependencies

Install

iti

~1kB inversion of control container for Typescript/Javascrith with a focus on async flow

  • 0.4.1
  • Source
  • npm
  • Socket score

Version published
Weekly downloads
2K
decreased by-25.38%
Maintainers
1
Weekly downloads
 
Created
Source

ITI Logo

CI Statu npm version gzip Mutation Score

Iti

1kB Depenendency Injection Framework for Typescript and Javascript with a unique feature that supports async flow

  • supports async(!): merges async code and constructor injection via plain async functions
  • non-invasive: does not require imported @decorators or framework extends in your application business logic
  • strongly typed: has great IDE autocomplete and compile time check. Without any manual type casting
  • lazy: initialises your app modules and containers on demand
  • split chunks: enables dynamic imports via a one liner thanks to a fully async core
  • React friendly: has useful React bindings to help you separate application business logic and a React view layer
  • starter friendly: works with starters like Create React App or Next.js unlike existing libraries
  • no Babel config: doesn't require reflect-metadata or decorators so there is no need to hack in decorator and "decoratorMetadata" support in to your build configs
  • tiny: less than 1kB

IoC is an amazing pattern and it should easy to adopt, fully support async and without hard to learn APIs or complex tooling requirements.

Iti relies on plain JS functions, objects and familiar patterns. API is simple so you can make a proof of concept integration in minutes.

It is an alternative to InversifyJS and microsoft/tsyringe for constructor injection.

At Packhelp we’ve refactored most of our 65K SLOC Editor app, that didn't have any IoC, to Iti in under 5 hours

Usage

// kitchen.ts
export class Oven {
  public pizzasInOven() {
    return 7
  }
  public async preheat() {}
}
export class Kitchen {
  constructor(public oven: Oven, public userManual: string) {}
}
// Application code is free of framework dependencies of decorators
// app.ts
import { makeRoot } from "iti"
import { Oven, Kitchen } from "./kitchen"

const node = makeRoot()
  .add({
    oven: () => new Oven(),
    userManual: async () => "Please preheat before use",
  })
  .add((ctx) => ({
    kitchen: async () => new Kitchen(ctx.oven, await ctx.userManual),
  }))

await node.get("kitchen") // Kitchen
// MyPizzaComponent.tsx
export const PizzaData = () => {
  const kitchen = useContainer().kitchen
  return <>Pizzaz In Oven: {kitchen.oven.pizzasInOven()}</>
}

Why another library?

The main reason is that existing libraries don’t support asynchronous code. Iti brings hassle free and fully typed way to use async code.

Secondly, existing libraries rely on decorators and reflect-metadata1. They couple your application business logic with a single framework and they tend to become unnecessarily complex. Also existing implementations will likely be incompatible with a TC39 proposal.

Also it is hard to use reflect-metadata with starters like CRA, Next.js etc. You need to eject or hack starters and it is far from ideal.

Short Manual

Reading

// Get a single instance
root.get("oven") // Creates a new Oven instance
root.get("oven") // Gets a cached Oven instance

await node.get("kitchen") // { kitchen: Kitchen } also cached
await node.containers.kitchen // same as above

// Get multiple instances at once
await root.getContainerSet(["oven", "userManual"]) // { userManual: '...', oven: Oven }
await root.getContainerSet((c) => [c.userManual, c.oven]) // same as above

// Subscribe to container changes
node.subscribeToContiner("oven", (oven) => {})
node.subscribeToContinerSet(["oven", "kitchen"], ({ oven, kitchen }) => {})
// prettier-ignore
node.subscribeToContinerSet((c) => [c.kitchen], ({ oven, kitchen }) => {})
node.on("containerUpdated", ({ key, newContainer }) => {})
node.on("containerUpserted", ({ key, newContainer }) => {})

Writing

let node1 = makeRoot()
  .add({
    userManual: "Please preheat before use",
    oven: () => new Oven(),
  })
  .upsert((containers, node) => ({
    userManual: "Works better when hot",
    preheatedOven: async () => {
      await containers.oven.preheat()
      return containers.oven
    },
  }))

// `add` is typesafe and a runtime safe method. Hence we've used `upsert`
try {
  node1.add({
    // @ts-expect-error
    userManual: "You shall not pass",
    // Type Error: (property) userManual: "You are overwriting this token. It is not safe. Use an unsafe `upsert` method"
  })
} catch (err) {
  err.message // Error Tokens already exist: ['userManual']
}

Patterns and tips

Lifecycle

Single Instance (a.k.a. Singleton)

let node = makeRoot().add({
  oven: () => new Oven(),
})
node.get("oven") === node.get("oven") // true

Transient

let node = makeRoot().add({
  oven: () => () => new Oven(),
})
node.get("oven") === node.get("oven") // false

Dynamic Imports

// ./kitchen/index.ts
export async function provideKitchenContainer() {
  const { Kitchen } = await import("./kitchen/kitchen")
  return {
    kitchen: () => new Kitchen(),
    oven: async () => {
      const { Oven } = await import("./kitchen/oven")
      const oven = new Oven()
      await oven.preheat()
      return oven
    },
  }
}
// ./index.ts
import { makeRoot } from "iti"
import { provideKitchenContainer } from "./kitchen"
let node = makeRoot().add({
  kitchen: async () => provideKitchenContainer(),
})

// Next line will load `./kitchen/kitchen` module
await node.containers.kitchen

// Next line will load `./kitchen/oven` module
await node.containers.kitchen.oven

Tip: Prefer callbacks over of strings (in progress)

If you use callback pattern across your app, you will be able to mass rename your containerKeys using typescript. With strings, you will have to manually go through the app. But even if you use string literals compiler will not compile until you fix your rename manually across the app.

const node = makeRoot().addNode({
  a: "A",
  b: "B",
})

await node.get((containerKeys) => containerKeys.a) // BEST!!!
await node.get("a") // it will work but...

Anti Patterns

in progress

Getting Started

The best way to get started is to check a CRA Pizza example

Initial wiring

import { makeRoot } from "../../src/library.new-root-container"

import { provideAContainer } from "./container.a"
import { provideBContainer } from "./container.b"
import { provideCContainer } from "./container.c"

export type MockAppNode = ReturnType<typeof getMainMockAppContainer>
export function getMainMockAppContainer() {
  return makeRoot()
    .add({ aCont: async () => provideAContainer() })
    .add((containers) => {
      return {
        bCont: async () => provideBContainer(await containers.aCont),
      }
    })
    .add((c) => {
      return {
        cCont: async () => provideCContainer(await c.aCont, await c.bCont, k),
      }
    })
}

Typescript

Iti has a great typescript support. All types are resolved automatically and check at compile time.

Autocomplete Autocomplete Autocomplete Autocomplete

Docs

Tokens

Containers

Containers are an important unit. If you replace them, users will be notified. In React it happens automatically

Events

const kitchenApp = new RootContainer((ctx) => ({
  // you can use tokens (`oven`, `kitchen`) here and later on
  oven: async () => ovenContainer(),
  kitchen: async () => kitchenContainer(await ctx.oven()),
}))

kitchenApp.on("containerCreated", (event) => {
  console.log(`event: 'containerCreated' ~~> token: '${event.key}'`)
  // `event.container` is also avaliable here
})

kitchenApp.on("containerRequested", (event) => {
  console.log(`event: 'containerRequested' ~~> token: '${event.key}' `)
})

kitchenApp.on("containerRemoved", (event) => {
  console.log(`event: 'containerRemoved' ~~> token: '${event.key}' `)
})

await kitchenApp.containers.kitchen

// event: 'containerRequested' ~~> token: 'kitchen'
// event: 'containerRequested' ~~> token: 'oven'
// event: 'containerCreated'   ~~> token: 'oven'
// event: 'containerCreated'   ~~> token: 'kitchen'

// Notice how oven was created before kitchen.
// This is because kitcen depends on oven

API documentation JS / TS

makeRoot Setting app root

import { makeRoot } from "../../library.root-container"
export function getMainMockAppContainer() {
  return makeRoot().add({ kitchen: () => new Kitchen(/* deps */) })
}

containers getter

let appRoot = getMainPizzaAppContainer()
let kitchen = await appRoot.containers.kitchen
kitchen.oven.pizzaCapacity // 4

getContainerSet

getContainerSetNew

upsert

When containers are updated React is updated too via hooks

Alternaitves

No async support

Existing libraries like inversify and others don’t support asynchronous code.

They either provide a promise to your constructor or require one to imperatively execute all potentially async code before the binding phase.

This is far from ideal.

Heavy use of decorators

Secondly, they rely on decorators and reflect-metadata

Decorators create unnecessary coupling of an application business logic with a framework. The whole idea of DI is to decouple the application business logic. Coupling classes with a DI framework is still coupling and turns DI into a service locator.

Also, decorator support is an experimental feature in Typescript and current implementation is not compatible with the TC39 proposal. This will probably cause problems for any non-trivial decorators and babel hacks.

In addition to that it is very hard to use reflect-metadata with starters like CRA, Next.js etc. To use reflect-metadata you need to tweak your compilers (babel, typescript, esbuild, swc etc.) configuration. So if you can’t use reflect-metadata you can’t use inversify or tsyringe.

Comparison with inversifyjs, tsyringe and others

Inversion of Control (IoC) is a great way to decouple code and the most popular pattern of IoC is dependency injection (DI) but it is not limited to one.

In JavaScript there is not way to create a dependency injection without mixing application business logic with a specific IoC library code or hacking a compiler (reflect-metadata).

inversifyjs and tsyringe use decorators and reflect-metada

import { injectable } from "tsyringe"

@injectable()
class Foo {
  constructor(private database: Database) {}
}

// some other file
import "reflect-metadata"
import { container } from "tsyringe"
import { Foo } from "./foo"

const instance = container.resolve(Foo)

typed-inject uses monkey-patching

import { createInjector } from "typed-inject"
function barFactory(foo: number) {
  return foo + 1
}
barFactory.inject = ["foo"] as const
class Baz {
  constructor(bar: number) {
    console.log(`bar is: ${bar}`)
  }
  static inject = ["bar"] as const
}

With Iti your application business logic is not mixed with the framework code

import type { Ingredients } from "./store.ingrediets"
import type { Oven } from "./store.oven"

export class Kitchen {
  constructor(private oven: Oven, private ingredients: Ingredients) {}
}

// provider / factory
import { IngredientsService } from "../services/ingredients-manager"
import { Kitchen } from "../stores/store.kitchen"
import { Oven } from "../stores/store.oven"

export async function provideKitchenContainer() {
  let oven = new Oven()
  let ingredients = await IngredientsService.buySomeIngredients()
  let kitchen = new Kitchen(oven, ingredients)

  return {
    oven: oven,
    ingredients: ingredients,
    kitchen: kitchen,
  }
}

Notable inspirations:

Questions and tips

Can I have multiple application containers?

Yes, no problem at all. If you want, they can even share tokens and hence instances!

Why getContainerSet is always async?

This is temporary(?) limitation to keep typescript happy and typescript types reasonable sane

Footnotes

  1. Kudos to typed-inject for finding a reasonable alternative to decorators and Reflect. Sadly, it doesn't support async and there are some other limits

Keywords

FAQs

Package last updated on 23 Jul 2022

Did you know?

Socket

Socket for GitHub automatically highlights issues in each pull request and monitors the health of all your open source dependencies. Discover the contents of your packages and block harmful activity before you install or update your dependencies.

Install

Related posts

SocketSocket SOC 2 Logo

Product

  • Package Alerts
  • Integrations
  • Docs
  • Pricing
  • FAQ
  • Roadmap
  • Changelog

Packages

npm

Stay in touch

Get open source security insights delivered straight into your inbox.


  • Terms
  • Privacy
  • Security

Made with ⚡️ by Socket Inc