Lightweight decoding and validation library.


Keywords
typescript, validation, validate, proof, schema, decoder
License
MIT
Install
npm install ts-prove@1.1.2

Documentation

TS Prove

A lightweight decoding and validation library.

GitHub Workflow Status Coverage Status Package Size

About

This library provides a set of composable functions for typesafe schema validation and data decoding. Because typescript is staticaly typed this can be particularly useful for message validation in, for example, a HTTP route handler by ensuring that the correct data is available.

Install

npm install ts-prove
yarn add ts-prove

Documentation

Example

In the example below, we construct a proof for the type Person<{ name: { first: string, last: string }, age: number }>. This proof is a callback which accepts some unknown data and returns either Failure<[string, unknown]> or Success<[null, Person]>

import P, { ProofType } from 'ts-prove'

const proveName = P.shape({
  first: P.string,
  last: P.optional(P.string),
})

const provePerson = P.shape({
  name: proveName,
  age: P.number,
})

type Person = ProofType<typeof provePerson>
// { name: { first: string, last?: string }, age: number }

provePerson({ name: { first: 'Bob' }, age: 'Incorect string value' }) // [string, unknown]
provePerson({ name: { first: 'Bob' }, age: 10 }) // [null, Person]

This could then be used to validate a payload.

import { isProved } from 'ts-prove'

export const createPerson = (req) => {
  const data = person(req.body)
  if (!isProved(data)) return req.send({ status: 500, body: data[0] })
  return db.createPerson(data[1])
}

db.createPerson = (p: Person) => void

Validation

Every proof is curried and accepts either a function of type (x: unknown) => true | string or some other value. When supplied with a function, a proof returns an instance of itself narrowed by the constraints of the callback. In the example bellow we use this feature to further narrow the number type to numbers betwen 10 and 19.

const teenage = P.number((x) => x > 10 || 'To young')((x) => x < 19 || 'To old')

teenage(9) // ['To young', unknown]
teenage(20) // ['To old', unknown]
teenage(13) // [null, number]

// This could then be used in a structured proof
const teenager = P.shape({ name: nameProof, age: teenage })

Proofs

All the proofs provided by this library are constructed using the function P<ValueType>(): Proof<ValueType>. We can use P to construct are own proofs by providing a return type as the type argument.

P.string = P<string>((x) => typeof x === 'string' || 'Expected string')

Here is a generic proof which does a deep equal comparison using lodash.isequal function.

import P from 'ts-prove'
import isEqual from 'lodash.isequal'

const equalProof = <T extends any>(type: T) => P<T>((input) => isEqual(type, input) | `Not equal`)

API

Type definitions

Type Definition
Sucess<T> [null, T]
Failure [string, unknown]
Check<T=unknown> (value: T) => true | string
Proof<R> <T *extends Check<R> | R*>(x: T): T extends Check<R> ? Proof<R> : Success<R> | Failure

P: Proof

Accepts a return type argument and a validation function that returns a failure of type string or boolean true.

  • string: Proof<string>
  • number: Proof<number>
  • boolean: Proof<boolean>
  • symbol: Proof<symbol>
  • null: Proof<null>
  • undefined: Proof<undefined>

Two utility proofs that always match but that resolve to different types.

  • any: Proof<any>
  • unknown: Proof<unknown>
  • array: <T extends Proof<any>>(proof: T): Proof<*ProofType<T>[]*>

    • Accepts a proof T as an argument and returns a proof for arrays containing type T.

    • P.array(P.string)
      // string[]
  • shape: <T extends { [x: string]: Proof<any> }>(proof: T): Proof<{ [Key in keyof T]: ProofType<T[Key]> }>

    • Accepts a key value object where every value is a proof and returns a proof for objects of that shape.

    • P.shape({ name: P.string, age: P.number })
      // { name: string, age: number }
  • optional: <T extends Proof<any>>(proof: T): Proof<*ProofType<T> | undefined*>

    • For use in shapes when you want the return type to indicate and optional value. Optional accepts a proof as an argument and returns union with undefined. For any other cases use or.

    • P.shape({ name: P.string, age: P.optional(P.number) })
      // { name: string, age?: number }
  • or: <T extends Proof<any>[]>(...proofs: T): Proof<ProofType<T[number]>>

    • Accepts any number of proofs as arguments and returns a union of thoses proofs if any one of them match a given value.

    • P.or(P.string, P.number, P.symbol)
      // string | number | symbol

Helper functions

  • isProved: <T extends any>(x : Failure | Success<T>): x is Success<T>

    • Type guard for the return type of a proof
  • check: (proof: Proof<any>): Check<unknown>

    • Used to convert a proof to a check

Type guard library coming soon...

This project follows the all-contributors specification. Contributions of any kind are welcome!