Angular Discussion Javascript jQuery NodeJS React Native React.js Vue.js


Yup is a JavaScript schema builder for value parsing and validation. Define a schema, transform a value to match, validate the shape of an existing value, or both. Yup schema are extremely expressive and allow modeling complex, interdependent validations, or value transformations.

Yup’s API is heavily inspired by Joi, but leaner and built with client-side validation as its primary use-case. Yup separates the parsing and validating functions into separate steps. cast() transforms data while validate checks that the input is the correct shape. Each can be performed together (such as HTML form validation) or seperately (such as deserializing trusted data from APIs).

Try it out:


npm install -S yup

Yup always relies on the Promise global object to handle asynchronous values as well as Set and Map. For browsers that do not support these, you’ll need to include a polyfill, such as core-js:

import 'core-js/es6/promise';
import 'core-js/es6/set';
import 'core-js/es6/map';

If you are using TypeScript installing the Yup typings is recommended

npm install -D @types/yup


You define and create schema objects. Schema objects are immutable, so each call of a method returns a new schema object. When using es module syntax, yup exports everything as a named export

 import * as yup from 'yup'; // for everything
// or
import { string, object } from 'yup'; // for only what you need
 let yup = require('yup');
let schema = yup.object().shape({
  name: yup.string().required(),
  age: yup.number().required().positive().integer(),
  email: yup.string().email(),
  website: yup.string().url(),
  createdOn: () {
    return new Date();
// check validity
    name: 'jimmy',
    age: 24,
  .then(function (valid) {
    valid; // => true
// you can try and type cast objects to the defined schema
  name: 'jimmy',
  age: '24',
  createdOn: '2014-09-23T19:25:25Z',
// => { name: 'jimmy', age: 24, createdOn: Date }

If you’re looking for an easily serializable DSL for yup schema, check out yup-ast

Using a custom locale dictionary

Allows you to customize the default messages used by Yup, when no message is provided with a validation test. If any message is missing in the custom dictionary the error message will default to Yup’s one.

import { setLocale } from 'yup';
  mixed: {
    default: 'Não é válido',
  number: {
    min: 'Deve ser maior que ${min}',
// now use Yup schemas AFTER you defined your custom dictionary
let schema = yup.object().shape({
  name: yup.string(),
  age: yup.number().min(18),
schema.validate({ name: 'jimmy', age: 11 }).catch(function (err) {; // => 'ValidationError'
  err.errors; // => ['Deve ser maior que 18']

If you need multi-language support, Yup has got you covered. The function setLocale accepts functions that can be used to generate error objects with translation keys and values. Just get this output and feed it into your favorite i18n library.

import { setLocale } from 'yup';
  // use constant translation keys for messages without values
  mixed: {
    default: 'field_invalid',
  // use functions to generate an error object that includes the value from the schema
  number: {
    min: ({ min }) => ({ key: 'field_too_short', values: { min } }),
    max: ({ max }) => ({ key: 'field_too_big', values: { max } }),
// now use Yup schemas AFTER you defined your custom dictionary
let schema = yup.object().shape({
  name: yup.string(),
  age: yup.number().min(18),
schema.validate({ name: 'jimmy', age: 11 }).catch(function (err) {; // => 'ValidationError'
  err.errors; // => [{ key: 'field_too_short', values: { min: 18 } }]



The module export.

let yup = require('yup');
yup.boolean; // also aliased as yup.bool;

yup.reach(schema: Schema, path: string, value?: object, context?: object): Schema

For nested schemas yup.reach will retrieve a nested schema based on the provided path.

For nested schemas that need to resolve dynamically, you can provide a value and optionally a context object.

let schema = object().shape({
  nested: object().shape({
    arr: array().of(object().shape({ num: number().max(4) })),
reach(schema, 'nested.arr.num');
reach(schema, 'nested.arr[].num');
reach(schema, 'nested.arr[1].num');
reach(schema, 'nested["arr"][1].num');

yup.addMethod(schemaType: Schema, name: string, method: ()=> Schema): void

Adds a new method to the core schema types. A friendlier convenience method for schemaType.prototype[name] = method.

yup.addMethod(, 'format', function (formats, parseStrict) {
  return this.transform(function (value, originalValue) {
    if (this.isType(value)) return value;
    value = Moment(originalValue, formats, parseStrict);
    return value.isValid() ? value.toDate() : new Date('');

yup.ref(path: string, options: { contextPrefix: string }): Ref

Creates a reference to another sibling or sibling descendant field. Refs are resolved at validation/cast time and supported where specified. Refs are evaluated in the proper order so that the ref value is resolved before the field using the ref (be careful of circular dependencies!).

let schema = object({
  baz: ref(''),
  foo: object({
    bar: string(),
  x: ref('$x'),
schema.cast({ foo: { bar: 'boom' } }, { context: { x: 5 } });
// => { baz: 'boom',  x: 5, foo: { bar: 'boom' } }

yup.lazy((value: any) => Schema): Lazy

Creates a schema that is evaluated at validation/cast time. Useful for creating recursive schema like Trees, for polymorphic fields and arrays.

CAUTION! When defining parent-child recursive object schema, you want to reset the default() to undefined on the child—otherwise the object will infinitely nest itself when you cast it!

let node = object({
  id: number(),
  child: yup.lazy(() => node.default(undefined)),
let renderable = yup.lazy((value) => {
  switch (typeof value) {
    case 'number':
      return number();
    case 'string':
      return string();
      return mixed();
let renderables = array().of(renderable);

ValidationError(errors: string | Array<string>, value: any, path: string)

Thrown on failed validations, with the following properties

  • name: “ValidationError”
  • path: a string, indicating where there error was thrown. path is empty at the root level.
  • errors: array of error messages
  • inner: in the case of aggregate errors, inner is an array of ValidationErrors throw earlier in the validation chain. When the abortEarly option is false this is where you can inspect each error thrown, alternatively, errors will have all of the messages from each inner error.


Creates a schema that matches all types. All types inherit from this base type

let schema = yup.mixed();
schema.isValid(undefined, function (valid) {
  valid; // => true

mixed.clone(): Schema

Creates a deep copy of the schema. Clone is used internally to return a new schema with every schema state change.

mixed.label(label: string): Schema

Overrides the key name which is used in error messages.

mixed.meta(metadata: object): Schema

Adds to a metadata object, useful for storing data with a schema, that doesn’t belong the cast object itself.

mixed.describe(): SchemaDescription

Collects schema details (like meta, labels, and active tests) into a serializable description object.

SchemaDescription {
  type: string,
  label: string,
  meta: object,
  tests: Array<{ name: string, params: object }>

mixed.concat(schema: Schema): Schema

Creates a new instance of the schema by combining two schemas. Only schemas of the same type can be concatenated.

mixed.validate(value: any, options?: object): Promise<any, ValidationError>

Returns the value (a cast value if isStrict is false) if the value is valid, and returns the errors otherwise. This method is asynchronous and returns a Promise object, that is fulfilled with the value, or rejected with a ValidationError.

The options argument is an object hash containing any schema options you may want to override (or specify for the first time).

Options = {
  strict: boolean = false;
  abortEarly: boolean = true;
  stripUnknown: boolean = false;
  recursive: boolean = true;
  context?: object;
  • strict: only validate the input, and skip any coercion or transformation
  • abortEarly: return from validation methods on the first error rather than after all validations run.
  • stripUnknown: remove unspecified keys from objects.
  • recursive: when false validations will not descend into nested schema (relevant for objects or arrays).
  • context: any context needed for validating schema conditions (see: when())
schema.validate({ name: 'jimmy', age: 24 }).then(function (value) {
  value; // => { name: 'jimmy',age: 24 }
schema.validate({ name: 'jimmy', age: 'hi' }).catch(function (err) {; // => 'ValidationError'
  err.errors; // => ['age must be a number']

mixed.validateSync(value: any, options?: object): any

Runs validatations synchronously if possible and returns the resulting value, or throws a ValidationError. Accepts all the same options as validate.

Synchronous validation only works if there are no configured async tests, e.g tests that return a Promise. For instance this will work:

let schema = number().test(
  "this isn't the number i want",
  (value) => value != 42,
schema.validateSync(23); // throws ValidationError

however this will not:

let schema = number().test('is-42', "this isn't the number i want", 
(value) =>
  Promise.resolve(value != 42),
schema.validateSync(42); // throws Error

mixed.validateAt(path: string, value: any, options?: object): Promise<any, ValidationError>

Validate a deeply nested path within the schema. Similar to how reach works, but uses the resulting schema as the subject for validation.

Note! The value here is the root value relative to the starting schema, not the value at the nested path.

let schema = object({
  foo: array().of(
      loose: boolean(),
      bar: string().when('loose', {
        is: true,
        otherwise: (s) => s.strict(),
let rootValue = {
  foo: [{ bar: 1 }, { bar: 1, loose: true }],
await schema.validateAt('foo[0].bar', rootValue); // => ValidationError: must be a string
await schema.validateAt('foo[1].bar', rootValue); // => '1'

mixed.validateSyncAt(path: string, value: any, options?: object): any

Same as validateAt but synchronous.

mixed.isValid(value: any, options?: object): Promise<boolean>

Returns true when the passed in value matches the schema. isValid is asynchronous and returns a Promise object.

Takes the same options as validate().

mixed.isValidSync(value: any, options?: object): boolean

Synchronously returns true when the passed in value matches the schema.

Takes the same options as validateSync() and has the same caveats around async tests.

mixed.cast(value: any, options = {}): any

Attempts to coerce the passed in value to a value that matches the schema. For example: '5' will cast to 5 when using the number() type. Failed casts generally return null, but may also return results like NaN and unexpected strings.

options parameter can be an object containing context. (For more info on context see mixed.validate)

mixed.isType(value: any): boolean

Runs a type check against the passed in value. It returns true if it matches, it does not cast the value. When nullable() is set null is considered a valid value of the type. You should use isType for all Schema type checks.

mixed.strict(isStrict: boolean = false): Schema

Sets the strict option to true. Strict schemas skip coercion and transformation attempts, validating the value “as is”.

mixed.strip(stripField: boolean = true): Schema

Marks a schema to be removed from an output object. Only works as a nested schema.

let schema = object({
  useThis: number(),
  notThis: string().strip(),
schema.cast({ notThis: 'foo', useThis: 4 }); // => { useThis: 4 }

mixed.withMutation(builder: (current: Schema) => void): void

First the legally required Rich Hickey quote:

If a tree falls in the woods, does it make a sound?

If a pure function mutates some local data in order to produce an immutable return value, is that ok?

withMutation allows you to mutate the schema in place, instead of the default behavior which clones before each change. Generally this isn’t necessary since the vast majority of schema changes happen during the initial declaration, and only happen once over the lifetime of the schema, so performance isn’t an issue. However certain mutations do occur at cast/validation time, (such as conditional schema using when()), or when instantiating a schema object.

  .shape({ key: string() })
  .withMutation((schema) => {
    return arrayOfObjectTests.forEach((test) => {

mixed.default(value: any): Schema

Sets a default value to use when the value is undefined. Defaults are created after transformations are executed, but before validations, to help ensure that safe defaults are specified. The default value will be cloned on each use, which can incur performance penalty for objects and arrays. To avoid this overhead you can also pass a function that returns a new default. Note that null is considered a separate non-empty value.

yup.object.default({ number: 5 }); // object will be cloned every time a default is needed
yup.object.default(() => ({ number: 5 })); // this is cheaper => new Date()); // also helpful for defaults that change over time

mixed.default(): Any

Calling default with no arguments will return the current default value

mixed.nullable(isNullable: boolean = true): Schema

Indicates that null is a valid value for the schema. Without nullable() null is treated as a different type and will fail isType() checks.

mixed.required(message?: string | function): Schema

Mark the schema as required. All field values apart from undefined and null meet this requirement.

mixed.notRequired(): Schema

Mark the schema as not required. Passing undefined as value will not fail validation.

mixed.defined(): Schema

Mark the schema as required but nullable. All field values apart from undefined meet this requirement.

mixed.typeError(message: string): Schema

Define an error message for failed type checks. The ${value} and ${type} interpolation can be used in the message argument.

mixed.oneOf(arrayOfValues: Array<any>, message?: string | function): Schema Alias: equals

Whitelist a set of values. Values added are automatically removed from any blacklist if they are in it. The ${values} interpolation can be used in the message argument.

Note that undefined does not fail this validator, even when undefined is not included in arrayOfValues. If you don’t want undefined to be a valid value, you can use mixed.required.

 let schema = yup.mixed().oneOf(['jimmy', 42]);
await schema.isValid(42); // => true
await schema.isValid('jimmy'); // => true
await schema.isValid(new Date()); // => false

mixed.notOneOf(arrayOfValues: Array<any>, message?: string | function)

Blacklist a set of values. Values added are automatically removed from any whitelist if they are in it. The ${values} interpolation can be used in the message argument.

 let schema = yup.mixed().notOneOf(['jimmy', 42]);
await schema.isValid(42); // => false
await schema.isValid(new Date()); // => true

mixed.when(keys: string | Array<string>, builder: object | (value, schema)=> Schema): Schema

Adjust the schema based on a sibling or sibling children fields. You can provide an object literal where the key is is value or a matcher function, then provides the true schema and/or otherwise for the failure condition.

is conditions are strictly compared (===) if you want to use a different form of equality you can provide a function like: is: (value) => value == true.

Like joi you can also prefix properties with $ to specify a property that is dependent on context passed in by validate() or isValidwhen conditions are additive.

let schema = object({
  isBig: boolean(),
  count: number()
    .when('isBig', {
      is: true, // alternatively: (val) => val == true
      then: yup.number().min(5),
      otherwise: yup.number().min(0),
    .when('$other', (other, schema) => (other === 4 ? schema.max(6) : schema)),
await schema.validate(value, { context: { other: 4 } });

You can also specify more than one dependent key, in which case each value will be spread as an argument.

let schema = object({
  isSpecial: boolean(),
  isBig: boolean(),
  count: number().when(['isBig', 'isSpecial'], {
    is: true, // alternatively: (isBig, isSpecial) => isBig && isSpecial
    then: yup.number().min(5),
    otherwise: yup.number().min(0),
await schema.validate({
  isBig: true,
  isSpecial: true,
  count: 10,

Alternatively you can provide a function that returns a schema (called with the value of the key and the current schema).

let schema = yup.object({
  isBig: yup.boolean(),
  count: yup.number().when('isBig', (isBig, schema) => {
    return isBig ? schema.min(5) : schema.min(0);
await schema.validate({ isBig: false, count: 4 });

mixed.test(name: string, message: string | function, test: function): Schema

Adds a test function to the validation chain. Tests are run after any object is cast. Many types have some tests built in, but you can create custom ones easily. In order to allow asynchronous custom validations all (or no) tests are run asynchronously. A consequence of this is that test execution order cannot be guaranteed.

All tests must provide a name, an error message and a validation function that must return true when the current value is valid and false or a ValidationError otherwise. To make a test async return a promise that resolves true or false or a ValidationError.

For the message argument you can provide a string which will interpolate certain values if specified using the ${param} syntax. By default all test messages are passed a path value which is valuable in nested schemas.

The test function is called with the current value. For more advanced validations you can use the alternate signature to provide more options (see below):

let jimmySchema = string().test(
  '${path} is not Jimmy',
  (value, context) => value === 'jimmy',
// or make it async by returning a promise
let asyncJimmySchema = string().test(
  '${path} is not Jimmy',
  async (value, context) => (await fetch('/is-jimmy/' + value)).responseText === 'true',
await schema.isValid('jimmy'); // => true
await schema.isValid('john'); // => false

Test functions are called with a special context, or this value, that exposes some useful metadata and functions. Older versions just expose the this context using function (), not arrow-func, but now it’s exposed too as a second argument of the test functions. It’s allow you decide which approach you prefer.

  • this.path: the string path of the current validation
  • this.schema: the resolved schema object that the test is running against.
  • this.options: the options object that validate() or isValid() was called with
  • this.parent: in the case of nested schema, this is the value of the parent object
  • this.originalValue: the original value that is being tested
  • this.createError(Object: { path: String, message: String, params: Object }): create and return a validation error. Useful for dynamically setting the pathparams, or more likely, the error message. If either option is omitted it will use the current path, or default message.

mixed.test(options: object): Schema

Alternative test(..) signature. options is an object containing some of the following options:

Options = {
  // unique name identifying the test
  name: string;
  // test function, determines schema validity
  test: (value: any) => boolean;
  // the validation error message
  message: string;
  // values passed to message for interpolation
  params: ?object;
  // mark the test as exclusive, meaning only one of the same can be active at once
  exclusive: boolean = false;

In the case of mixing exclusive and non-exclusive tests the following logic is used. If a non-exclusive test is added to a schema with an exclusive test of the same name the exclusive test is removed and further tests of the same name will be stacked.

If an exclusive test is added to a schema with non-exclusive tests of the same name the previous tests are removed and further tests of the same name will replace each other.

let max = 64;
let schema = yup.mixed().test({
  name: 'max',
  exclusive: true,
  params: { max },
  message: '${path} must be less than ${max} characters',
  test: (value) => value == null || value.length <= max,

mixed.transform((currentValue: any, originalValue: any) => any): Schema

Adds a transformation to the transform chain. Transformations are central to the casting process, default transforms for each type coerce values to the specific type (as verified by isType()). transforms are run before validations and only applied when the schema is not marked as strict (the default). Some types have built in transformations.

Transformations are useful for arbitrarily altering how the object is cast, however, you should take care not to mutate the passed in value. Transforms are run sequentially so each value represents the current state of the cast, you can use the originalValue param if you need to work on the raw initial value.

let schema = string().transform(function (value, originalvalue) {
  return this.isType(value) && value !== null ? value.toUpperCase() : value;
schema.cast('jimmy'); // => 'JIMMY'

Each types will handle basic coercion of values to the proper type for you, but occasionally you may want to adjust or refine the default behavior. For example, if you wanted to use a different date parsing strategy than the default one you could do that with a transform.

module.exports = function (formats = 'MMM dd, yyyy') {
  return date().transform(function (value, originalvalue) {
    // check to see if the previous transform already parsed the date
    if (this.isType(value)) return value;
    // the default coercion failed so let's try it with Moment.js instead
    value = Moment(originalValue, formats);
    // if it's valid return the date object, otherwise return an `InvalidDate`
    return value.isValid() ? value.toDate() : new Date('');


Define a string schema. Supports all the same methods as mixed.

let schema = yup.string(); await schema.isValid('hello'); // => true

By default, the cast logic of string is to call toString on the value if it exists. empty values are not coerced (use ensure() to coerce empty values to empty strings).

Failed casts return the input value.

string.required(message?: string | function): Schema

The same as the mixed() schema required, except that empty strings are also considered ‘missing’ values.

string.length(limit: number | Ref, message?: string | function): Schema

Set a required length for the string value. The ${length} interpolation can be used in the message argument

string.min(limit: number | Ref, message?: string | function): Schema

Set a minimum length limit for the string value. The ${min} interpolation can be used in the message argument

string.max(limit: number | Ref, message?: string | function): Schema

Set a maximum length limit for the string value. The ${max} interpolation can be used in the message argument

string.matches(regex: Regex, message?: string | function): Schema

Provide an arbitrary regex to match the value against.

let schema = string().matches(/(hi|bye)/); await schema.isValid('hi'); // => trueawait schema.isValid('nope'); // => false

string.matches(regex: Regex, options: { message: string, excludeEmptyString: bool }): Schema

An alternate signature for string.matches with an options object. excludeEmptyString, when true, short circuits the regex test when the value is an empty string

let schema = string().matches(/(hi|bye)/, { excludeEmptyString: true }); 
await schema.isValid(''); // => true string | function): Schema

Validates the value as an email address via a regex.

string.url(message?: string | function): Schema

Validates the value as a valid URL via a regex.

string.uuid(message?: string | function): Schema

Validates the value as a valid UUID via a regex.

string.ensure(): Schema

Transforms undefined and null values to an empty string along with setting the default to an empty string.

string.trim(message?: string | function): Schema

Transforms string values by removing leading and trailing whitespace. If strict() is set it will only validate that the value is trimmed.

string.lowercase(message?: string | function): Schema

Transforms the string value to lowercase. If strict() is set it will only validate that the value is lowercase.

string.uppercase(message?: string | function): Schema

Transforms the string value to uppercase. If strict() is set it will only validate that the value is uppercase.


Define a number schema. Supports all the same methods as mixed.

let schema = yup.number(); 
await schema.isValid(10); // => true

The default cast logic of number is: parseFloat.

Failed casts return NaN.

number.min(limit: number | Ref, message?: string | function): Schema

Set the minimum value allowed. The ${min} interpolation can be used in the message argument.

number.max(limit: number | Ref, message?: string | function): Schema

Set the maximum value allowed. The ${max} interpolation can be used in the message argument.

number.lessThan(max: number | Ref, message?: string | function): Schema

Value must be less than max. The ${less} interpolation can be used in the message argument.

number.moreThan(min: number | Ref, message?: string | function): Schema

Value must be strictly greater than min. The ${more} interpolation can be used in the message argument.

number.positive(message?: string | function): Schema

Value must be a positive number.

number.negative(message?: string | function): Schema

Value must be a negative number.

number.integer(message?: string | function): Schema

Validates that a number is an integer.

number.truncate(): Schema

Transformation that coerces the value to an integer by stripping off the digits to the right of the decimal point.

number.round(type: 'floor' | 'ceil' | 'trunc' | 'round' = 'round'): Schema

Adjusts the value via the specified method of Math (defaults to ’round’).


Define a boolean schema. Supports all the same methods as mixed.

let schema = yup.boolean(); await schema.isValid(true); // => true


Define a Date schema. By default ISO date strings will parse correctly, for more robust parsing options see the extending schema types at the end of the readme. Supports all the same methods as mixed.

let schema =; await schema.isValid(new Date()); // => true

The default cast logic of date is pass the value to the Date constructor, failing that, it will attempt to parse the date as an ISO date string.

Failed casts return an invalid Date.

date.min(limit: Date | string | Ref, message?: string | function): Schema

Set the minimum date allowed. When a string is provided it will attempt to cast to a date first and use the result as the limit.

date.max(limit: Date | string | Ref, message?: string | function): Schema

Set the maximum date allowed, When a string is provided it will attempt to cast to a date first and use the result as the limit.


Define an array schema. Arrays can be typed or not, When specifying the element type, cast and isValid will apply to the elements as well. Options passed into isValid are passed also passed to child schemas. Supports all the same methods as mixed.

let schema = yup.array().of(yup.number().min(2));
await schema.isValid([2, 3]); // => true
await schema.isValid([1, -24]); // => false
schema.cast(['2', '3']); // => [2, 3]

You can also pass a subtype schema to the array constructor as a convenience.

array().of(yup.number());// orarray(yup.number());

The default cast behavior for array is: JSON.parse

Failed casts return: null;

array.of(type: Schema): Schema

Specify the schema of array elements. of() is optional and when omitted the array schema will not validate its contents.

array.required(message?: string | function): Schema

The same as the mixed() schema required, except that empty arrays are also considered ‘missing’ values.

array.min(limit: number | Ref, message?: string | function): Schema

Set a minimum length limit for the array. The ${min} interpolation can be used in the message argument.

array.max(limit: number | Ref, message?: string | function): Schema

Set a maximum length limit for the array. The ${max} interpolation can be used in the message argument.

array.ensure(): Schema

Ensures that the value is an array, by setting the default to [] and transforming null and undefined values to an empty array as well. Any non-empty, non-array value will be wrapped in an array.

array().ensure().cast(null); // => []
array().ensure().cast(1); // => [1]
array().ensure().cast([1]); // => [1]

array.compact(rejector: (value) => boolean): Schema

Removes falsey values from the array. Providing a rejecter function lets you specify the rejection criteria yourself.

array().compact().cast(['', 1, 0, 4, false, null]); // => [1, 4]
  .compact(function (v) {
    return v == null;
  .cast(['', 1, 0, 4, false, null]); // => ['', 1, 0, 4, false]


Define an object schema. Options passed into isValid are also passed to child schemas. Supports all the same methods as mixed.

  name: string().required(),
  age: number().required().positive().integer(),
  email: string().email(),
  website: string().url(),

You can also pass a shape to the object constructor as a convenience.

  num: number(),
// or
  num: number(),

The default cast behavior for object is: JSON.parse

Failed casts return: null;

Object schema defaults

Object schema come with a default value already set, which “builds” out the object shape, a sets any defaults for fields:

const schema = object({
  name: string().default(''),
schema.default(); // -> { name: '' }

This may be a bit suprising, but is generally very helpful since it allows large, nested schema to create default values that fill out the whole shape and not just the root object. There is one gotcha! though. For nested object schema that are optional but include non optional fields may fail in unexpected ways:

const schema = object({
  id: string().required(),
  names: object({
    first: string().required(),
schema.isValid({ id: 1 }); // false! names.first is required

This is because yup casts the input object before running validation which will produce:

{ id: '1', names: { first: undefined }}

During the validation phase names exists, and is validated, finding names.first missing. If you wish to avoid this behavior do one of the following:

  • Set the nested default to undefined: names.default(undefined)
  • mark it nullable and default to null: names.nullable().default(null)

object.shape(fields: object, noSortEdges?: Array<[string, string]>): Schema

Define the keys of the object and the schemas for said keys.

Note that you can chain shape method, which acts like object extends, for example:

  a: string(),
  b: number(),
  b: string(),
  c: number(),

would be exactly the same as:

  a: string(),
  b: string(),
  c: number(),

object.from(fromKey: string, toKey: string, alias: boolean = false): Schema

Transforms the specified key to a new key. If alias is true then the old key will be left.

let schema = object({
  myProp: mixed(),
  Other: mixed(),
  .from('prop', 'myProp')
  .from('other', 'Other', true);
schema.cast({ prop: 5, other: 6 }); // => { myProp: 5, other: 6, Other: 6 }

object.noUnknown(onlyKnownKeys: boolean = true, message?: string | function): Schema

Validate that the object value only contains keys specified in shape, pass false as the first argument to disable the check. Restricting keys to known, also enables stripUnknown option, when not in strict mode.

object.camelCase(): Schema

Transforms all object keys to camelCase

object.constantCase(): Schema

Transforms all object keys to CONSTANT_CASE.

Extending Schema Types

The simplest way to extend an existing type is just to cache a configured schema and use that through your application.

let yup = require('yup');
let parseFormats = ['MMM dd, yyy'];
let invalidDate = new Date('');
module.exports = (value, originalValue) {
  if (this.isType(value)) return value;
  // the default coercion transform failed so let's try it with Moment instead
  value = Moment(originalValue, parseFormats);
  return value.isValid() ? value.toDate() : invalidDate;

Alternatively, each schema is a normal JavaScript constructor function that you can mutate or delegate to using the normal patterns. Generally you should not inherit from mixed unless you know what you are doing, better to think of it as an abstract class. The other types are fair game though.

You should keep in mind some basic guidelines when extending schemas:

  • never mutate an existing schema, always clone() and then mutate the new one before returning it. Built-in methods like test and transform take care of this for you, so you can safely use them (see below) without worrying
  • transforms should never mutate the value passed in, and should return an invalid object when one exists (NaNInvalidDate, etc) instead of null for bad values.
  • by the time validations run the value is guaranteed to be the correct type, however if nullable is set then null is a valid value for that type, so don’t assume that a property or method exists on the value.

Adjust core Types

let invalidDate = new Date('');
function parseDateFromFormats(formats, parseStrict) {
  return this.transform(function (value, originalValue) {
    if (this.isType(value)) return value;
    value = Moment(originalValue, formats, parseStrict);
    return value.isValid() ? value.toDate() : invalidDate;

Creating new Types

Yup schema use the common constructor pattern for modeling inheritance. You can use any utility or pattern that works with that pattern. The below demonstrates using the ES6 class syntax since it’s less verbose, but you absolutely aren’t required to use it.

let DateSchema =;
let invalidDate = new Date(''); // our failed to coerce value
class MomentDateSchemaType extends DateSchema {
  constructor() {
    this._validFormats = [];
    this.withMutation(() => {
      this.transform(function (value, originalvalue) {
        if (this.isType(value))
          // we have a valid value
          return value;
        return Moment(originalValue, this._validFormats, true);
  _typeCheck(value) {
    return (
      super._typeCheck(value) || (moment.isMoment(value) && value.isValid())
  format(formats) {
    if (!formats) throw new Error('must enter a valid format');
    let next = this.clone();
    next._validFormats = {}.concat(formats);
let schema = new MomentDateSchemaType();
schema.format('YYYY-MM-DD').cast('It is 2012-05-25'); // => Fri May 25 2012 00:00:00 GMT-0400 (Eastern Daylight Time)

TypeScript Support

If you are using TypeScript installing the Yup typings is recommended:

npm install -D @types/yup

You can now infer a TypeScript type alias using the exported InferType. Given the following Yup schema:

import * as yup from 'yup';
const personSchema = yup.object({
  firstName: yup
    // Here we use `defined` instead of `required` to more closely align with
    // TypeScript. Both will have the same effect on the resulting type by
    // excluding `undefined`, but `required` will also disallow other values
    // such as empty strings.
  nickName: yup
  gender: yup
    // Note `as const`: this types the array as `["male", "female", "other"]`
    // instead of `string[]`.
    .oneOf(['male', 'female', 'other'] as const)
  email: yup
  birthDate: yup
    .min(new Date(1900, 0, 1)),

You can derive the TypeScript type as follows:

type Person = yup.InferType<typeof personSchema>;

Which is equivalent to the following TypeScript type alias:

type Person = {
  firstName: string;
  nickName: string | null;
  gender: "male" | "female" | "other";
  email?: string | null | undefined;
  birthDate?: Date | null | undefined;

Making the following objects valid both for TypeScript and Yup validation:

const minimalPerson: Person = {
    firstName: "Matt",
    nickName: null,
    gender: "male"
const fullPerson: Person = {
    firstName: "Matt",
    nickName: "The Hammer",
    gender: "male",
    email: "",
    birthDate: new Date(1976, 9, 5)

You can also go the other direction, specifying an interface and ensuring that a schema matches it:

type Person = {
  firstName: string;
// ✔️ compiles
const goodPersonSchema: yup.ObjectSchema<Person> = yup.object({
  firstName: yup.string().defined()
// ❌ errors:
// "Type 'number | undefined' is not assignable to type 'string'."
const badPersonSchema: yup.ObjectSchema<Person> = yup.object({
  firstName: yup.number()

TypeScript setting

For yup.InferType<T> to work correctly with required and nullable types you have to set strict: true or strictNullChecks: true in your tsconfig.json.

You may also like...

Leave a Reply

Your email address will not be published. Required fields are marked *