✨ A superior validation solution for your NestJS application ✨
Ecosystem
Package | About |
---|---|
nestjs-zod | A tools for integrating Zod into your NestJS application |
nestjs-zod-prisma | Generate Zod schemas from your Prisma schema |
Core library features
createZodDto
- create DTO classes from Zod schemasZodValidationPipe
- validatebody
/query
/params
using Zod DTOsZodGuard
- guard routes by validatingbody
/query
/params
(it can be useful when you want to do that before other guards)UseZodGuard
- alias for@UseGuards(new ZodGuard(source, schema))
ZodValidationException
- BadRequestException extended with Zod errorszodToOpenAPI
- create OpenAPI declarations from Zod schemas- OpenAPI support
@nestjs/swagger
integration using the patchzodToOpenAPI
- generate highly accurate Swagger Schema- Zod DTOs can be used in any
@nestjs/swagger
decorator
- Extended Zod schemas for NestJS (
nestjs-zod/z
)dateString
for dates (supports casting toDate
)password
for passwords (more complex string rules + OpenAPI conversion)
- Customization - change exception format easily
- Useful helpers for client side error handling (
nestjs-zod/frontend
)
Installation
yarn add nestjs-zod zod
Peer dependencies:
zod
->= 3.14.3
@nestjs/common
->= 8.0.0
(required on server side)@nestjs/core
->= 8.0.0
(required on server side)@nestjs/swagger
->= 5.0.0
(only when usingpatchNestJsSwagger
)
All peer dependencies are marked as optional for better client side usage, but you need to install required ones when using nestjs-zod
on server side.
Navigation
- Writing Zod schemas
- Creating DTO from Zod schema
- Using ZodValidationPipe
- Using ZodGuard
- Validation Exceptions
- Using ZodSerializerInterceptor
- Extended Zod
- OpenAPI (Swagger) support
Writing Zod schemas
Extended Zod and Swagger integration are bound to the internal API, so even the patch updates can cause errors.
For that reason, nestjs-zod
uses specific zod
version inside and re-exports it under /z
scope:
import { z, ZodString, ZodError } from 'nestjs-zod/z'
const CredentialsSchema = z.object({
username: z.string(),
password: z.string(),
})
Zod's classes and types are re-exported too, but under /z
scope for more clarity:
import { ZodString, ZodError, ZodIssue } from 'nestjs-zod/z'
Creating DTO from Zod schema
import { createZodDto } from 'nestjs-zod'
import { z } from 'nestjs-zod/z'
const CredentialsSchema = z.object({
username: z.string(),
password: z.string(),
})
// class is required for using DTO as a type
class CredentialsDto extends createZodDto(CredentialsSchema) {}
Using DTO
DTO does two things:
- Provides a schema for
ZodValidationPipe
- Provides a type from Zod schema for you
@Controller('auth')
class AuthController {
// with global ZodValidationPipe (recommended)
async signIn(@Body() credentials: CredentialsDto) {}
async signIn(@Param() signInParams: SignInParamsDto) {}
async signIn(@Query() signInQuery: SignInQueryDto) {}
// with route-level ZodValidationPipe
@UsePipes(ZodValidationPipe)
async signIn(@Body() credentials: CredentialsDto) {}
}
// with controller-level ZodValidationPipe
@UsePipes(ZodValidationPipe)
@Controller('auth')
class AuthController {
async signIn(@Body() credentials: CredentialsDto) {}
}
Using standalone (without server-side dependencies)
import { createZodDto } from 'nestjs-zod/dto'
Using ZodValidationPipe
The validation pipe uses your Zod schema to parse data from parameter decorator.
When the data is invalid - it throws ZodValidationException.
Globally (recommended)
import { ZodValidationPipe } from 'nestjs-zod'
import { APP_PIPE } from '@nestjs/core'
@Module({
providers: [
{
provide: APP_PIPE,
useClass: ZodValidationPipe,
},
],
})
export class AppModule {}
Locally
import { ZodValidationPipe } from 'nestjs-zod'
// controller-level
@UsePipes(ZodValidationPipe)
class AuthController {}
class AuthController {
// route-level
@UsePipes(ZodValidationPipe)
async signIn() {}
}
Also, you can instantly pass a Schema or DTO:
import { ZodValidationPipe } from 'nestjs-zod'
import { UserDto, UserSchema } from './auth.contracts'
// using schema
@UsePipes(new ZodValidationPipe(UserSchema))
// using DTO
@UsePipes(new ZodValidationPipe(UserDto))
class AuthController {}
class AuthController {
// the same applies to route-level
async signIn() {}
}
Creating custom validation pipe
import { createZodValidationPipe } from 'nestjs-zod'
const MyZodValidationPipe = createZodValidationPipe({
// provide custom validation exception factory
createValidationException: (error: ZodError) =>
new BadRequestException('Ooops'),
})
Using ZodGuard
Sometimes, we need to validate user input before specific Guards. We can't use Validation Pipe since NestJS Pipes are always executed after Guards.
The solution is ZodGuard
. It works just like ZodValidationPipe
, except for that is doesn't transform the input.
It has 2 syntax forms:
@UseGuards(new ZodGuard('body', CredentialsSchema))
@UseZodGuard('body', CredentialsSchema)
Parameters:
- The source -
'body' | 'query' | 'params'
- Zod Schema or DTO (just like
ZodValidationPipe
)
When the data is invalid - it throws ZodValidationException.
import { ZodGuard } from 'nestjs-zod'
// controller-level
@UseZodGuard('body', CredentialsSchema)
@UseZodGuard('params', CredentialsDto)
class MyController {}
class MyController {
// route-level
@UseZodGuard('query', CredentialsSchema)
@UseZodGuard('body', CredentialsDto)
async signIn() {}
}
Creating custom guard
import { createZodGuard } from 'nestjs-zod'
const MyZodGuard = createZodGuard({
// provide custom validation exception factory
createValidationException: (error: ZodError) =>
new BadRequestException('Ooops'),
})
Create validation from scratch
If you don't like ZodGuard
and ZodValidationPipe
, you can use validate
function:
import { validate } from 'nestjs-zod'
validate(wrongThing, UserDto, (zodError) => new MyException(zodError)) // throws MyException
const validatedUser = validate(
user,
UserDto,
(zodError) => new MyException(zodError)
) // returns typed value when succeed
Validation Exceptions
The default server response on validation error looks like that:
{
"statusCode": 400,
"message": "Validation failed",
"errors": [
{
"code": "too_small",
"minimum": 8,
"type": "string",
"inclusive": true,
"message": "String must contain at least 8 character(s)",
"path": ["password"]
}
]
}
The reason of this structure is default ZodValidationException
.
You can customize the exception by creating custom nestjs-zod
entities using the factories:
You can create ZodValidationException
manually by providing ZodError
:
const exception = new ZodValidationException(error)
Also, ZodValidationException
has an additional API for better usage in NestJS Exception Filters:
@Catch(ZodValidationException)
export class ZodValidationExceptionFilter implements ExceptionFilter {
catch(exception: ZodValidationException) {
exception.getZodError() // -> ZodError
}
}
Using ZodSerializerInterceptor for output validation
To ensure that a response conforms to a certain shape, you may use the ZodSerializerInterceptor
interceptor.
This would be especially useful in prevent accidental data leaks.
This is similar to NestJs' @ClassSerializerInterceptor
feature here
@ZodSerializerInterceptor
in application root
Include @Module({
...
providers: [
...,
{ provide: APP_INTERCEPTOR, useClass: ZodSerializerInterceptor },
],
})
export class AppModule {}
@ZodResponseDto
to define the shape of the response for endpoint in controller
Use const UserSchema = z.object({ username: string() })
export class UserDto extends createZodDto(UserSchema) {}
@Controller('user')
export class UserController {
constructor(private readonly userService: UserService) {}
@ZodResponseDto(UserDto)
getUser(id: number) {
return this.userService.findOne(id) // --> The native service method returns { username: string, password: string by default }
}
}
In the above example, despite the userService.findOne
method returns password
, the password
property will be stripped out thanks to the @ZodResponseDto
decorator.
Extended Zod
As you learned in Writing Zod Schemas section, nestjs-zod
provides a special version of Zod. It helps you to validate the user input more accurately by using our custom schemas and methods.
ZodDateString
In HTTP, we always accept Dates as strings. But default Zod doesn't have methods to validate such type of strings. ZodDateString
was created to address this issue.
// 1. Expect user input to be a "string" type
// 2. Expect user input to be a valid date (by using new Date)
z.dateString()
// Cast to Date instance
// (use it on end of the chain, but before "describe")
z.dateString().cast()
// Expect string in "full-date" format from RFC3339
z.dateString().format('date')
// [default format]
// Expect string in "date-time" format from RFC3339
z.dateString().format('date-time')
// Expect date to be the past
z.dateString().past()
// Expect date to be the future
z.dateString().future()
// Expect year to be greater or equal to 2000
z.dateString().minYear(2000)
// Expect year to be less or equal to 2025
z.dateString().maxYear(2025)
// Expect day to be a week day
z.dateString().weekDay()
// Expect year to be a weekend
z.dateString().weekend()
Valid date
format examples:
2022-05-15
Valid date-time
format examples:
2022-05-02:08:33Z
2022-05-02:08:33.000Z
2022-05-02:08:33+00:00
2022-05-02:08:33-00:00
2022-05-02:08:33.000+00:00
Errors:
-
invalid_date_string
- invalid date -
invalid_date_string_format
- wrong formatPayload:
expected
-'date' | 'date-time'
-
invalid_date_string_direction
- not past/futurePayload:
expected
-'past' | 'future'
-
invalid_date_string_day
- not weekDay/weekendPayload:
expected
-'weekDay' | 'weekend'
-
too_small
withtype === 'date_string_year'
-
too_big
withtype === 'date_string_year'
ZodPassword
ZodPassword
is a string-like type, just like the ZodDateString
. As you might have guessed, it's intended to help you with password schemas definition.
Also, ZodPassword
has a more accurate OpenAPI conversion, comparing to regular .string()
: it has password
format and generated RegExp string for pattern
.
// Expect user input to be a "string" type
z.password()
// Expect password length to be greater or equal to 8
z.password().min(8)
// Expect password length to be less or equal to 100
z.password().max(100)
// Expect password to have at least one digit
z.password().atLeastOne('digit')
// Expect password to have at least one lowercase letter
z.password().atLeastOne('lowercase')
// Expect password to have at least one uppercase letter
z.password().atLeastOne('uppercase')
// Expect password to have at least one special symbol
z.password().atLeastOne('special')
Errors:
invalid_password_no_digit
invalid_password_no_lowercase
invalid_password_no_uppercase
invalid_password_no_special
too_small
withtype === 'password'
too_big
withtype === 'password'
Json Schema
Created for
nestjs-zod-prisma
z.json()
"from" function
Created for custom schemas in
nestjs-zod-prisma
Just returns the same Schema
z.from(MySchema)
Extended Zod Errors
Currently, we use custom
error code due to some Zod limitations (errorMap
priorities)
Therefore, the error details is located inside params
property:
const error = {
code: 'custom',
message: 'Invalid date, expected it to be the past',
params: {
isNestJsZod: true,
code: 'invalid_date_string_direction',
// payload is always located here in a flat view
expected: 'past',
},
path: ['date'],
}
Working with errors on the client side
Optionally, you can install nestjs-zod
on the client side.
The library provides you a /frontend
scope, that can be used to detect custom NestJS Zod issues and process them the way you want.
import { isNestJsZodIssue, NestJsZodIssue, ZodIssue } from 'nestjs-zod/frontend'
function mapToFormErrors(issues: ZodIssue[]) {
for (const issue of issues) {
if (isNestJsZodIssue(issue)) {
// issue is NestJsZodIssue
}
}
}
⚠️ If you usezod
in your client-side application, and you want to installnestjs-zod
too, it may be better to completely switch tonestjs-zod
to prevent issues caused by mismatch betweenzod
versions.nestjs-zod/frontend
doesn't usezod
at the runtime, but it uses its types.
OpenAPI (Swagger) support
Setup
Prerequisites:
@nestjs/swagger
with version^5.0.0
installed
Apply a patch:
import { patchNestJsSwagger } from 'nestjs-zod'
patchNestJsSwagger()
Then follow the Nest.js' Swagger Module Guide.
Writing more Swagger-compatible schemas
Use .describe()
method to add Swagger description:
import { z } from 'nestjs-zod/z'
const CredentialsSchema = z.object({
username: z.string().describe('This is an username'),
password: z.string().describe('This is a password'),
})
Using zodToOpenAPI
You can convert any Zod schema to an OpenAPI JSON object:
import { zodToOpenAPI } from 'nestjs-zod'
import { z } from 'nestjs-zod/z'
const SignUpSchema = z.object({
username: z.string().min(8).max(20),
password: z.string().min(8).max(20),
sex: z
.enum(['male', 'female', 'nonbinary'])
.describe('We respect your gender choice'),
social: z.record(z.string().url()),
birthDate: z.dateString().past(),
})
const openapi = zodToOpenAPI(SignUpSchema)
The output will be the following:
{
"type": "object",
"properties": {
"username": {
"type": "string",
"minLength": 8,
"maxLength": 20
},
"password": {
"type": "string",
"minLength": 8,
"maxLength": 20
},
"sex": {
"description": "We respect your gender choice",
"type": "string",
"enum": ["male", "female", "nonbinary"]
},
"social": {
"type": "object",
"additionalProperties": {
"type": "string",
"format": "uri"
}
},
"birthDate": {
"type": "string",
"format": "date-time"
}
},
"required": ["username", "password", "sex", "social", "birthDate"]
}
Credits
-
zod-dto
nestjs-zod
includes a lot of refactored code fromzod-dto
. -
zod-nestjs and zod-openapi
These libraries bring some new features compared tozod-dto
.
nestjs-zod
has used them too.