TypeMap
Uniform Syntax, Compiler and Mapping for Runtime Types
Install
$ npm install @sinclair/typemap --save
Usage
Parse and Compile Types from TypeScript Syntax (Example)
import { Compile } from '@sinclair/typemap'
const result = Compile('string | null').Parse('Hello TypeMap')
// │ │ │
// │ │ └─── Parse Value
// │ │
// │ └── Ast: TUnion<[TString, TNull]>
// │
// └── const result: string | null = 'Hello TypeMap'
Overview
TypeMap is an syntax frontend and compiler backend for the TypeBox, Valibot and Zod libraries. It provides a common TypeScript syntax for type construction, a runtime compiler for high-performance validation and provides type translation from one library to another.
TypeMap is written to be an advanced type translation system for the TypeBox project. It is built to integrate and accelerate remote type libraries on Json Schema compatible infrastructure as well as to integrate TypeBox into remote type library infrastructure via reverse type remapping. This project also provides high-performance validation for frameworks that orientate around the Standard Schema TypeScript interface.
License: MIT
Contents
Example
Use TypeScript syntax to create types for TypeBox, Valibot and Zod (Example)
import { TypeBox, Valibot, Zod } from '@sinclair/typemap'
// Syntax
const S = `{
x: number,
y: number,
z: number
}`
// Runtime Types
const T = TypeBox(S) // const T: TObject<{
// x: TNumber,
// y: TNumber,
// z: TNumber
// }>
const V = Valibot(S) // const V: ObjectSchema<{
// x: NumberSchema<...>,
// y: NumberSchema<...>,
// z: NumberSchema<...>
// }, ...>
const Z = Zod(S) // const Z: ZodObject<{
// x: ZodNumber,
// y: ZodNumber,
// z: ZodNumber
// }, ...>
Translate TypeBox, Valibot and Zod types (Example)
import { TypeBox, Valibot, Zod } from '@sinclair/typemap'
// Syntax -> Zod -> Valibot -> TypeBox
const T = TypeBox(Valibot(Zod(`{
x: number,
y: number,
z: number
}`)))
Compile Valibot and Zod types on TypeBox validation infrastructure. (Example)
import { Compile } from '@sinclair/typemap'
import z from 'zod'
// Zod Type
const Z = z.object({ // const Z: ZodObject<{
x: z.number(), // x: ZodNumber,
y: z.number(), // y: ZodNumber,
z: z.number(), // z: ZodNumber
}) // }>
// Remap and Compile
const C = Compile(Z) // const C: Validator<TObject<{
// x: TNumber,
// y: TNumber,
// z: TNumber
// }>>
// High Throughout Validation
const R = C.Check({ // Iterations: 10_000_000
x: 1, //
y: 2, // Zod : 4000ms (approx)
z: 3 // TypeMap : 40ms (approx)
})
Mapping
TypeMap is primarily a mapping system intended for type translation. It provides mapping functions per library which is used to translate remote types into types specific to that library. All mapping functions make a best attempt to retain semantics from each library. If no translation is possible, these functions return a never representation specific to the library being mapped.
TypeBox
Use the TypeBox function to translate types and syntax into TypeBox types.
import { TypeBox } from '@sinclair/typemap'
const S = TypeBox('string[]') // const S: TArray<TString> (Syntax)
const T = TypeBox(t.Number()) // const T: TNumber (TypeBox)
const V = TypeBox(v.string()) // const V: TString (Valibot)
const Z = TypeBox(z.boolean()) // const Z: TBoolean (Zod)
Valibot
Use the Valibot function to translate types and syntax into Valibot types.
import { Valibot } from '@sinclair/typemap'
const S = Valibot('string[]') // const S: v.ArraySchema<...> (Syntax)
const T = Valibot(t.Number()) // const T: v.NumberSchema (TypeBox)
const V = Valibot(v.string()) // const V: v.StringSchema (Valibot)
const Z = Valibot(z.boolean()) // const Z: v.BooleanSchema (Zod)
Zod
Use the Zod function to translate types and syntax into Zod types.
import { Zod } from '@sinclair/typemap'
const S = Zod('string[]') // const S: z.ZodArray<...> (Syntax)
const T = Zod(t.Number()) // const T: z.ZodNumber (TypeBox)
const V = Zod(v.string()) // const V: z.ZodString (Valibot)
const Z = Zod(z.boolean()) // const Z: z.ZodBoolean (Zod)
Syntax
TypeMap provides a TypeScript syntax parser that can be used to create library types. TypeScript parsing is implemented at runtime as well as in the TypeScript type system. It offers a convenient means of creating cross library types without having to author across multiple type builder API.
Types
Syntax types can be created by passing a string parameter to any library mapping function. TypeMap supports most TypeScript annotation syntax. If the string contains a syntax error, the function will return a never type. (Example)
import { TypeBox } from '@sinclair/typemap'
const T = TypeBox('{ x: 1 }') // const T: TObject<{
// x: TLiteral<1>
// }>
const S = TypeBox('!!!') // const S: TNever - Syntax Error
Options
Options can be passed on the last parameter of a type. TypeMap will translate known Json Schema keywords into appropriate runtime representations if possible. (Example)
import { TypeBox, Zod } from '@sinclair/typemap'
const T = TypeBox('string', { // const T: TString = {
format: 'email' // type: 'string',
}) // format: 'email'
// }
const S = Zod('{ x: number }', { // const S = z.object({
additionalProperties: false // x: z.number()
}) // }).strict()
Parameters
Types can be parameterized to accept exterior types. (Example)
import { Valibot, Zod } from '@sinclair/typemap'
const T = Valibot('number') // const T: NumberSchema
// Parameter T auto remapped to target library
const S = Zod({ T }, `{ x: T }`) // const S: ZodObject<{
// x: ZodNumber
// }, { ... }>
Generics
Use parameterized types with functions to create generic types (Example)
import { TypeBox, Valibot, Zod } from '@sinclair/typemap'
// Generic Type
const Vector = <T extends object>(T: T) => TypeBox({ T }, `{
x: T,
y: T,
z: T
}`)
// Instanced Types
const T = Vector(Valibot('number')) // const T: TObject<{
// x: TNumber,
// y: TNumber,
// z: TNumber,
// }>
const S = Vector(Zod('string')) // const S: TObject<{
// x: TString,
// y: TString,
// z: TString,
// }>
Performance
TypeMap implements type level syntax parsing in the TypeScript type system. This feature puts a lot of additional demand on the TypeScript language service above and beyond the compute required to infer the underlying type library. One should approach this feature with some mindfulness to the computational overhead required parse and infer types from syntax. The following link can be used to test infer performance.
(Example)
import { TypeBox, Static } from '@sinclair/typemap'
// Application Types
const Country = TypeBox(`
'New Zealand' |
'Australia' |
'South Korea' |
'Japan' |
'Canada' |
'United States' |
'Norway' |
'Sweden' |
({} & string)
`)
const Address = TypeBox({ Country }, `{
street: string
city: string
postcode: string
country: Country
}`)
const UserId = TypeBox('string', { format: 'uuid' })
const Email = TypeBox('string', { format: 'email' })
const User = TypeBox({ UserId, Email, Address }, `{
userId: UserId,
username: string
email: Email
address: Address
created: Date
}`)
const Role = TypeBox(`
'adminstrator' |
'editor' |
'publisher'
`)
const SystemUser = TypeBox({ User, Role }, `User & {
roles: Role[]
}`)
// Performance Test
function performance(user: Static<typeof SystemUser>) {
user // use '.' to test infer performance
}
Static
Use Static to infer for library and syntax types
import { type Static } from '@sinclair/typemap'
const T = t.Number() // TypeBox
const V = v.string() // Valibot
const Z = z.boolean() // Zod
const S = 'string[]' // Syntax
type S = Static<typeof S> // string[]
type T = Static<typeof T> // number
type V = Static<typeof V> // string
type Z = Static<typeof Z> // boolean
TreeShake
TypeMap exports the top-level TypeBox, Valibot, and Zod functions, which broadly translate any type. However, applications are most likely going to be interested in translating between two libraries (at most) and in one specific direction (e.g., Zod to TypeBox). TypeMap provides specific functions that perform only a particular translation. Using these specific mapping functions allows bundlers to tree-shake unused type libraries.
import { TypeBoxFromZod } from '@sinclair/typemap' // Bundle TypeBox | Zod, Tree Shake Valibot
import * as z from 'zod'
const T = TypeBoxFromZod(z.object({ // const T: TObject<{
x: z.number(), // x: TNumber;
y: z.number(), // y: TNumber;
z: z.number() // z: TNumber;
})) // }>
Compile
Use the Compile function to compile Zod and Valibot on TypeBox validation infrastructure. (Example)
import { Compile, Zod } from '@sinclair/typemap'
// Compile Validator From Zod
const validator = Compile(Zod(`{
x: number,
y: number,
z: number
}`))
const R1 = validator.Check({ x: 1, y: 2, z: 3 })
// Standard Schema
//
// ... which should have been named 'Standard Validator'
const R2 = validator['~standard'].validate({ x: 1, y: 2, z: 3 })
Benchmark
This project manages a small benchmark that compares validation performance using Zod, Valibot, and TypeBox validators. For more comprehensive community benchmarks, refer to the runtime-type-benchmarks project.
Test
Benchmarks are run for the following type.
type T = { x: number, y: string, z: boolean }
Results
Results show the approximate elapsed time to complete the given iterations
┌─────────┬────────────────┬────────────────────┬────────────┬────────────┐
│ (index) │ library │ using │ iterations │ elapsed │
├─────────┼────────────────┼────────────────────┼────────────┼────────────┤
│ 0 │ 'valibot ' │ 'valibot ' │ 10000000 │ '1534 ms ' │
│ 1 │ 'valibot ' │ 'typebox:value ' │ 10000000 │ '1377 ms ' │
│ 2 │ 'valibot ' │ 'typebox:compile ' │ 10000000 │ '46 ms ' │
└─────────┴────────────────┴────────────────────┴────────────┴────────────┘
┌─────────┬────────────────┬────────────────────┬────────────┬────────────┐
│ (index) │ library │ using │ iterations │ elapsed │
├─────────┼────────────────┼────────────────────┼────────────┼────────────┤
│ 0 │ 'zod ' │ 'zod ' │ 10000000 │ '4669 ms ' │
│ 1 │ 'zod ' │ 'typebox:value ' │ 10000000 │ '1359 ms ' │
│ 2 │ 'zod ' │ 'typebox:compile ' │ 10000000 │ '47 ms ' │
└─────────┴────────────────┴────────────────────┴────────────┴────────────┘
Contribute
This project is open to community contributions. Please ensure you submit an open issue before creating a pull request. TypeBox and associated projects preference open community discussion before accepting new features.