Runcheck

A lib for js/typescript runtime type checks with autofix support. Runcheck has the goal of being very lightweight and fast ⚡. Because of that, it has only around 2.9kb Gzipped (at v0.30), has no dependencies and is tree-shakeable!

Obs: Runcheck is in Beta and it's api can still change

Benchmarks

One of the goals of runcheck is to be blazing fast. Here are some benchmarks:

Click to see the benchmarks

runcheck dist is the same version as runcheck in the benchmarks but bundled

Installation

pnpm add runcheck

Basic types:

runcheck type	ts type equivalent
rc_string	string
rc_number	number
rc_boolean	boolean
rc_any	any
rc_null	null
rc_undefined	undefined
rc_date	Date
rc_intanceof(instance: T)	Classes typecheck in general
rc_literals(...literals: T[])	Type literal in general like hello, true, 1
rc_union(...types: T[])	Union types in general like string | 1
rc_array<T>(type: T)	T[]
rc_tuple<T>(...types: T[])	[T, T]
rc_intersection(...types: T[])	Intersection types like {a:string} & {b:string}

Array types:

Array loose check

You can also use rc_loose_array to reject the wrong elements of an array and return the valid ones.

const shape = rc_loose_array(rc_string)

const input = ['hello', 1, 'world']

const result = rc_parse(input, shape)

// result.data will be ['hello', 'world']
// result.warnings will return the warnings about the invalid elements

Checking unique values

With the rc_array or rc_loose_array type you can also use the unique option to check if the array has no duplicated values.

const shape = rc_array(rc_string, { unique: true })

For arrays of objects, you can provide a string to unique option to check if the array items has no duplicated values of a specific property.

const shape = rc_array(rc_object({ id: rc_string }), { unique: 'id' })

You can also provide a function to unique option to check if the array items has no duplicated values based on a custom function return.

const shape = rc_array(
  rc_object({ id: rc_string, meta_id: rc_string.optional() }),
  {
    unique: (item) => item.meta_id || item.id,
  },
)

Object types:

rc_object

const shape = rc_object({
  name: rc_string,
  age: rc_number,
  isCool: rc_boolean,
  // nested objects
  address: {
    street: rc_string,
    number: rc_number,
  },
})

The rc_object will allow extra properties but, any extra propertie will be striped in parsing. To allow extra in parsing properties, use rc_extends_obj.

Marking optional keys

Optional keys can be marked with the optionalKey() method.

const shape = rc_object({
  name: rc_string.optionalKey(),
  age: rc_number,
  isCool: rc_boolean,
})

/*
infered type will be:
{
  name?: string | undefined,
  age: number,
  isCool: boolean,
}

instead of:
{
  name: string | undefined,
  age: number,
  isCool: boolean,
}
*/

rc_obj_strict

The same as rc_object but, any extra properties will be throw an error in parsing.

rc_obj_merge

Allow to merge two rc_object types. Example:

const shape = rc_obj_merge(
  rc_object({
    name: rc_string,
    age: rc_number,
    isCool: rc_boolean,
  }),
  rc_object({
    address: rc_string,
    phone: rc_string,
  }),
)

rc_record

Validates only the values of a object, equivalent to Record<string, T> in typescript.

const shape = rc_record(rc_number)

// shape type is `Record<string, number>`

// `rc_record` also accepts the following options:
const shape = rc_record(rc_number, {
  checkKey: (key) => key !== 'a', // Check if the key is valid
  looseCheck: true, // If true, the invalid keys will be striped
})

rc_loose_record

Validates only the values of a object, equivalent to Record<string, T> in typescript. But, it will reject invalid keys and return the valid ones.

const shape = rc_loose_record(rc_number)

Parsing

import { rc_parse } from 'runcheck'

const input = JSON.parse(jsonInput)

const parseResult = rc_parse(input, rc_array(rc_string))

if (parseResult.error) {
  throw new Error(parseResult.errors.join('\n'))
  // Errors are a array of strings
}

const result = parseResult.data
// Do something with result

You can also use rc_parser to create a reusable parser.

import { rc_parser } from 'runcheck'

const parser = rc_parser(rc_array(rc_string))

const parseResult = parser(jsonInput)
const parseResult2 = parser(jsonInput2)

Strict parsing

Use the strict option to disable autofix and fallback

const parseResult = rc_parse(
  input,
  // fallback will be ignored
  rc_array(rc_string).withFallback([]),
  {
    strict: true,
  },
)

Type assertion

Use rc_is_valid and rc_validator to do a simple type assertion.

import { rc_is_valid } from 'runcheck'

const input = JSON.parse(jsonInput)

if (rc_is_valid(input, rc_array(rc_string))) {
  // input will be inferred by ts as `string[]`
}

Type assertion in a parse result

Use rc_assert_is_valid to do a simple type assertion in a parse result.

import { rc_assert_is_valid } from 'runcheck'

const input = JSON.parse(jsonInput)

const result = rc_parse(input, rc_array(rc_string))

rc_assert_is_valid(result)
// will throw an error if the result is invalid, otherwise will narrow the result type to a valid result

Loose parsing

You can now use the result methods unwrapOr and unwrapOrNull for more convenient loose parsing:

import { rc_parse } from 'runcheck'

const input = JSON.parse(jsonInput)

// Using unwrapOr method
const result = rc_parse(input, rc_array(rc_string)).unwrapOr([])
// will fallback to [] if the input is invalid

// Using unwrapOrNull method
const result2 = rc_parse(input, rc_array(rc_string)).unwrapOrNull()
// will fallback to null if the input is invalid

Strict parsing

You can use the unwrap() method directly on the parse result:

import { rc_parse, RcValidationError } from 'runcheck'

const input = JSON.parse(jsonInput)

try {
  const result = rc_parse(input, rc_array(rc_string)).unwrap()
  // result will be the parsed data or throw RcValidationError
} catch (error) {
  if (error instanceof RcValidationError) {
    // handle error
  }
}

Autofixing and fallback values in parsing

Values can be autofixed and fallback values can be provided for parsing. The checks will pass but the result will return warnings messages.

type SuccessResult = {
  error: false
  data: T
  warnings: string[] | false
}

Fallback

Use the method rc_[type].withFallback(fallback) to provide a fallback value if the input is not valid.

const input = 'hello'

const result = rc_parse(input, rc_string.withFallback('world'))

AutoFix

You can also use rc_[type].autoFix() to automatically fix the input if it is not valid.

const input = 1

const result = rc_parse(
  input,
  rc_string.autoFix((input) => input.toString()),
)

There are also some predefined autofixed types that you can import:

import { rc_string_autofix, rc_boolean_autofix } from 'runcheck/autofixable'

// use like any other type

Performing custom checks

You can also use rc_[type].where(customCheckFunction) to perform custom checks.

const input = 1

const positiveNumberType = rc_number.where((input) => input > 0)

Infer types from schemas

You can use RcInferType<typeof schema> to infer the types from a schema.

const schema = rc_object({
  name: rc_string,
  age: rc_number,
  isCool: rc_boolean,
})

export type Person = RcInferType<typeof schema>

You can also use the RcPrettyInferType<typeof schema> to get a more readable type.

Type modifiers

You can use also modiers like rc_string.optional() to extend the rc types:

runcheck modifier	ts type equivalent
rc_[type].optional()	T | undefined
rc_[type].orNull()	T | null
rc_[type].orNullish()	T | null | undefined

Recursive types

You can use rc_recursive to create recursive types. But the types can't be inferred in this case. So you need to provide the type manually.

type MenuTree = {
  name: string
  children: MenuTree[]
}

// the type should be provided manually to the variable in this case
const menuTreeSchema: RcType<MenuTree[]> = rc_array(
  rc_object({
    name: rc_string,
    // you can safely autorefence the schema here
    children: rc_recursive(() => menuTreeSchema),
  }),
)

const result = rc_parse(input, menuTreeSchema)

Transform types

You can use rc_transform to validate an input and transform it to another data.

const input = 'hello'

const result = rc_parse(
  input,
  rc_transform(rc_string, (input) => input.length),
)

Use the outputSchema option to create a type that validates both the input and the output of the transform. So if the input matches the outputSchema the transform will be ignored.

const input = 'hello'

const schema = rc_transform(rc_string, (input) => input.length, {
  outputSchema: rc_number,
})

const result = rc_parse(input, schema)

if (result.ok) {
  // this will be valid too
  const transformedResult = rc_parse(result.data, schema)
}

// Be carefull: `outputSchema` will be used only if the input type is invalid

const schema = rc_transform(
  rc_union(rc_string, rc_number),
  (input) => String(input).toUperCase(),
  {
    // this will be ignored because has an equivalent type to the input
    outputSchema: rc_string,
  },
)

// use a more strict input type to avoid this

const schema = rc_transform(
  rc_union(rc_string, rc_number).where((input) => isNotUperCase(input)),
  (input) => String(input).toUperCase(),
  {
    outputSchema: rc_string.where((input) => isUperCase(input)),
  },
)

Tranformed types which result can be validated with same schema

You may want to create a transformed type which result can be validated with the same schema. For this you can use the rc_narrow type. Example:

const stringOrArrayOfStrings = rc_union(rc_string, rc_array(rc_string))

const schema = rc_narrow(stringOrArrayOfStrings, (input) =>
  Array.isArray(input) ? input : [input],
)

const result = rc_parse('hello', schema)

if (result.ok) {
  // the schema can safely be used to validate the result too
  const transformedResult = rc_parse(result.data, schema)
}

Default types

You can use rc_default to provide a default value if the input is undefined.

const input = {
  name: 'John',
}

const result = rc_parse(
  input,
  rc_object({
    name: rc_string,
    age: rc_default(rc_number, 20),
  }),
)

if (result.ok) {
  result.data.age // = 20
}

If you need to use default in nullish values you can use rc_nullish_default.

Advanced object types

rc_get_from_key_as_fallback

Allows to rename a key in a object. Example:

const shape = rc_object({
  // name will use the value of oldName if name is not present in input
  // which will rename `oldName` to `name` in the result
  name: rc_get_from_key_as_fallback('oldName', rc_string),
  age: rc_number,
  isCool: rc_boolean,
})

Snake case normalization

you can use rc_object with the normalizeKeysFrom option to normalize the keys of a object to snake case.

const shape = rc_object(
  {
    name: rc_string,
    age: rc_number,
    isCool: rc_boolean,
  },
  { normalizeKeysFrom: 'snake_case' },
)

rc_parse({ name: 'John', age: 20, is_cool: true }, shape) // will not return an error and will normalize the response to { name: 'John', age: 20, isCool: true }

rc_get_obj_schema

Allows to get a subset of a object schema. Example:

const shape = rc_object({
  name: rc_string,
  age: rc_number,
  isCool: rc_boolean,
})

const nameSchema = rc_get_obj_schema(shape).name

rc_obj_extends

Don't strip unchecked keys from the result. Example:

const shape = rc_object({
  name: rc_string,
})

const result = rc_parse(
  { name: 'John', age: 20, is_cool: true },
  rc_obj_extends(shape),
)

// keys `age` and `is_cool` will be present in the result
result.data // { name: 'John', age: 20, is_cool: true }

rc_obj_pick

Allows to pick a subset of a object schema. Example:

const shape = rc_object({
  name: rc_string,
  age: rc_number,
  isCool: rc_boolean,
})

const nameSchema = rc_obj_pick(shape, ['name'])

rc_obj_omit

Allows to omit a subset of a object schema. Example:

const shape = rc_object({
  name: rc_string,
  age: rc_number,
  isCool: rc_boolean,
})

const baseSchema = rc_obj_omit(shape, ['isCool'])

rc_obj_builder

Creates a rc_object from a type. This gives better error messages and autocompletion.

type SchemaType = {
  level1: {
    level2: {
      level3: {
        level4: {
          level5: number
        }
      }
    }
  }
  optionalObj?: {
    a: string
  }
  objOrNull: null | {
    a: string
  }
  objOrNullish:
    | null
    | undefined
    | {
        a: string
      }
}

const schema = rc_obj_builder<SchemaType>()({
  level1: {
    level2: {
      level3: {
        level4: {
          level5: rc_string,
          // better error here
        },
      },
    },
  },
  optionalObj: [
    'optional',
    {
      a: rc_string,
      // better error here and autocompletion :)
    },
  ],
  objOrNull: [
    'null_or',
    {
      a: rc_string,
    },
  ],
  objOrNullish: [
    'nullish_or',
    {
      a: rc_string,
    },
  ],
})

rc_discriminated_union

Creates a discriminated union type with faster check performance compared to rc_union.

const networkState = rc_discriminated_union('state', {
  loading: {},
  success: {
    response: rc_string,
  },
  error: {
    code: rc_number,
  },
})

const result = rc_unwrap(
  rc_parse({ state: 'success', response: 'hello' }, networkState),
)
// result will be inferred as:
// | { state: 'loading' }
// | { state: 'success', response: string }
// | { state: 'error', code: number }

rc_array_filter_from_schema

Creates a two passes array validation. The first will validate the items against the filter schema and filter the item. The second will perform the type check against the filtered items.

const schema = rc_array_filter_from_schema(
  // 1 validate the items against a filter schema
  rc_object({
    deleted: rc_boolean,
  }),
  // Then filter the items based on the filter schema result
  (item) => !item.deleted,

  // 2 validate the filtered items
  rc_object({
    value: rc_string,
  }),
)

const result = rc_parse(
  [
    { deleted: false, value: 'hello' },
    { deleted: true, value: 'world' },
  ],
  schema,
)

// result.value === [{ value: 'hello' }]