Skip to content

Latest commit

 

History

History
240 lines (175 loc) · 6.54 KB

README.md

File metadata and controls

240 lines (175 loc) · 6.54 KB

io-ts-transformer

TypeScript transformer which converts TypeScript types to io-ts entities

Requirement

TypeScript >= 3.5.2

io-ts >= 2.x+

fp-ts >= 2.0.0

Installation

To install this package run the following command:

npm i io-ts-transformer io-ts fp-ts

About this project

Io-ts is an awesome library that makes possible runtime type validation in TypeScript. It has many advantages, but there is one fundamental flaw: it makes us duplicate information in our projects, which leads to many other problems. Let's look at this simple structure:

type User = {
    name: string
    age: number
}

As information about this type will be lost when compiling from TypeScript to JavaScript, we need to describe an io-ts entity that would be able to validate our data at runtime.

import * as t from 'io-ts'

const user = t.type({
  name: t.string,
  age: t.number
})

This approach has significant disadvantages:

  1. Duplicating the same information in different forms takes time for the developer.

  2. If you need to change TypeScript data type, you must not forget to change the io-ts model too. Otherwise, you can get an elusive bug or vulnerability in the system security.

The solution is automatic compile-time transformation of TypeScript types into io-ts models. Io-ts-transformer does this for you.

Package structure

This package exports 2 functions. One is buildDecoder which is used in TypeScript code to convert TypeScript types into io-ts entities, while the other is a TypeScript custom transformer which is used to compile the buildDecoder function correctly.

Io-ts-transformer already can:

  1. Transform almost all TypeScript types into io-ts models:
  • null, undefined, void, unknown
  • string, boolean and number literals
  • string, boolean and number types
  • arrays, tuples, records, objects and functions
  • type unions and intersections
  • interfaces and recursive types
  1. Compute expressions passed into it.

For example, this expression

buildDecoder<Omit<{ foo: 'bar', bar: number } & { data: string }, 'bar'>>()

will be converted into

import * as t from 'io-ts'

t.type({ foo: t.literal('bar'), data: t.string })
  1. Inject existing io-ts models into a TypeScript type. You can use FromIoTs type provided by io-ts-transformer to specify variables to inject. Then io-ts models, built by io-ts-transformer, will reference the variables passed.
import { withFallback } from 'io-ts-types/lib/withFallback'
import { buildDecoder, FromIoTs } from 'io-ts-transformer'
import * as t from 'io-ts'

const options = t.type({
  logLevel: withFallback(t.string, 'info')
})

type Config = {
  folder: string
  options: FromIoTs<typeof options>
}

// const config = t.type({ folder: t.string, options: options })
const config = buildDecoder<Config>()

Don't forget wrapping your models in t.recursion, if you use buildDecoder inside them, and the type passed includes FromIoTs recursive reference to this model. For example, this code will produce a runtime error:

import { buildDecoder, FromIoTs } from 'io-ts-transformer'
import * as t from 'io-ts'

const foo = t.type({ bar: buildDecoder<T>() })

type T = { foo?: FromIoTs<typeof foo> }

// Will produce a runtime error
const decodeResult = foo.decode({ bar: {} })

To avoid this you should wrap foo model in t.recursion:

import { buildDecoder, FromIoTs } from 'io-ts-transformer'
import * as t from 'io-ts'

const foo = t.recursion('foo', () => t.type({ bar: buildDecoder<T>() }))

type T = { foo?: FromIoTs<typeof foo> }

// Now it will work as intended
const decodeResult = foo.decode({ bar: {} })

Io-ts-transformer can't do

  1. Transform classes.

  2. Work with dynamic type parameters, i.e. buildDecoder<T>() in the following code will be converted into t.void as default io-ts entity:

import { buildDecoder } from 'io-ts-transformer'

function convertEntity<T>(entity: T) {
  return buildDecoder<T>()
}

How to use buildDecoder

import { buildDecoder } from 'io-ts-transformer'

type User = {
  name: string
  age: number
}

// const usersDecoder = t.array(t.type({ name: t.string, age: t.number }))
const usersDecoder = buildDecoder<Array<User>>()

How to use the custom transformer

Unfortunately, TypeScript itself does not currently provide any easy way to use custom transformers (See microsoft/TypeScript#14419). The followings are the example usage of the custom transformer.

webpack (with ts-loader or awesome-typescript-loader)

// webpack.config.js
const ioTsTransformer = require('io-ts-transformer/transformer').default

module.exports = {
  // ...
  module: {
    rules: [
      {
        test: /\.ts$/,
        loader: 'ts-loader', // or 'awesome-typescript-loader'
        options: {
          // make sure not to set `transpileOnly: true` here, otherwise it will not work
          getCustomTransformers: program => ({
              before: [
                  ioTsTransformer(program)
              ]
          })
        }
      }
    ]
  }
}

Rollup (with rollup-plugin-typescript2)

// rollup.config.js
import resolve from 'rollup-plugin-node-resolve'
import typescript from 'rollup-plugin-typescript2'
import ioTsTransformer from 'io-ts-transformer/transformer'

export default {
  // ...
  plugins: [
    resolve(),
    typescript({ transformers: [service => ({
      before: [ ioTsTransformer(service.getProgram()) ],
      after: []
    })] })
  ]
}

ttypescript

See ttypescript's README for how to use this with module bundlers such as webpack or Rollup.

// tsconfig.json
{
  "compilerOptions": {
    // ...
    "plugins": [
      { "transform": "io-ts-transformer/transformer" }
    ]
  },
  // ...
}

Author's note

This package is implemented as a diploma project, and it probably would be nice to insert a couple of reviews into diploma report from people, who had time to try this package in their own projects. If you want, you can write me at awerlogus@yandex.ru or in Telegram. Don't forget to report bugs you have found. I wish you a pleasant use. With love awerlogus.

Thanks to:

  • Kimamula for his project, where I took some code and readme file parts.
  • My friend Vladimir for checking the grammar of my writings.

License

MIT