Prisma Types Generator

1. What is it?
2. Usage
3. Annotations
4. Example
5. Principles

What is it?

Generates ConnectDTO, CreateDTO, UpdateDTO, DTO, and Entity classes for models in your Prisma Schema. This is useful if you want to leverage OpenAPI in your NestJS application - but also helps with GraphQL resources as well). NestJS Swagger requires input parameters in controllers to be described through classes because it leverages TypeScript's emitted metadata and Reflection to generate models/components for the OpenAPI spec. It does the same for response models/components on your controller methods. Output files are formatted with prettier using project config. Every output model is configurable, so you can choose what models to generate. These classes can also be used with the built-in ValidationPipe and Serialization.

Usage?

npm install --save-dev prisma-types-generator

generator types {
  provider                        = "prisma-types-generator"
  output                          = "../src/generated/types"
  outputToNestJsResourceStructure = "false"
  flatResourceStructure           = "false"
  exportRelationModifierClasses   = "true"
  reExport                        = "false"
  createDtoPrefix                 = "Create"
  updateDtoPrefix                 = "Update"
  dtoSuffix                       = "Dto"
  entityPrefix                    = ""
  entitySuffix                    = ""
  classValidation                 = "false"
  fileNamingStyle                 = "camel"
  noDependencies                  = "false"
  excludeEntity                   = "false"
  excludePlainDto                 = "false"
  excludeCreateDto                = "false"
  excludeUpdateDto                = "false"
  excludeConnectDto               = "false"
  definiteAssignmentAssertion     = "false"
}

Parameters

All parameters are optional.

• [output]: (default: "../src/generated/types") - output path relative to your schema.prisma file
• [outputToNestJsResourceStructure]: (default: "false") - writes dtos and entities to subfolders aligned with NestJS CRUD generator. Resource module name is derived from lower-cased model name in schema.prisma
• [flatResourceStructure]: (default: "false") - If outputToNestJsResourceStructure is true, subfolders dtos and entities are created within the resource folder. Setting this to true will flatten the hierarchy.
• [exportRelationModifierClasses]: (default: "true") - Should extra classes generated for relationship field operations on DTOs be exported?
• [reExport]: (default: false) - Should an index.ts be created for every folder?
• [createDtoPrefix]: (default: "Create") - phrase to prefix every CreateDTO class with
• [updateDtoPrefix]: (default: "Update") - phrase to prefix every UpdateDTO class with
• [dtoSuffix]: (default: "Dto") - phrase to suffix every CreateDTO and UpdateDTO class with
• [entityPrefix]: (default: "") - phrase to prefix every Entity class with
• [entitySuffix]: (default: "") - phrase to suffix every Entity class with
• [excludeEntity]: (default: "false") - exclude Entity generation
• [excludePlainDto]: (default: "false") - exclude DTO generation
• [excludeCreateDto]: (default: "false") - exclude CreateDTO generation
• [excludeUpdateDto]: (default: "false") - exclude UpdateDTO generation
• [excludeConnectDto]: (default: "false") - exclude ConnectDTO generation
• [fileNamingStyle]: (default: "camel") - How to name generated files. Valid choices are "camel", "pascal", "kebab" and "snake".
• [classValidation]: (default: "false") - Add validation decorators from class-validator. Not compatible with noDependencies = "true" and outputType = "interface".
• [noDependencies]: (default: "false") - Any imports and decorators that are specific to NestJS and Prisma are omitted, such that there are no references to external dependencies. This is useful if you want to generate appropriate DTOs for the frontend.
• [outputType]: (default: "class") - Output the DTOs as class or as interface. interface should only be used to generate DTOs for the frontend.
• [definiteAssignmentAssertion]: (default: "false") - Use definite assignment assertion operator (!) on optional fields. Enable this if you are using TypeScript's strictPropertyInitialization option.

Annotations

Annotations provide additional information to help this generator understand your intentions. They are applied as tripple slash comments to a field node in your Prisma Schema. You can apply multiple annotations to the same field.

model Post {
  /// @DtoCreateOptional
  /// @DtoUpdateHidden
  createdAt   DateTime @default(now())
}

• @DtoReadOnly - omits field in CreateDTO and UpdateDTO
• @DtoEntityHidden - omits field in Entity
• @DtoCreateOptional - adds field optionally to CreateDTO - useful for fields that would otherwise be omitted (e.g. @id, @updatedAt)
• @DtoUpdateOptional- adds field optionally to UpdateDTO - useful for fields that would otherwise be omitted (e.g. @id, @updatedAt)
• @DtoRelationRequired - marks relation required in Entity although it's optional in PrismaSchema - useful when you don't want (SQL) ON DELETE CASCADE behavior - but your logical data schema sees this relation as required
  (Note: becomes obsolete once referentialActions are released and stable)
• @DtoRelationCanCreateOnCreate - adds create option on a relation field in the generated CreateDTO - useful when you want to allow to create related model instances
• @DtoRelationCanConnectOnCreate - adds connect option on a relation field in the generated CreateDTO - useful when you want/need to connect to an existing related instance
• @DtoRelationCanCreateOnUpdate - adds create option on a relation field in the generated UpdateDTO - useful when you want to allow to create related model instances
• @DtoRelationCanConnectOnUpdate - adds connect option on a relation field in the generated UpdateDTO - useful when you want/need to connect to an existing related instance

Schema Object annotations

With @nestjs/swagger, you can generate an API specification from code. Routes, request bodies, query parameters, etc., are annotated with special decorators. Properties can be annotated with the @ApiProperty() decorator to add schema object information. They are partially added at runtime, which will then include type, nullable, etc. But additional information, such as description, need to be added manually.

If using a generator like this, any custom @ApiProperty() annotation would be overridden when updating the DTOs. To enhance a field with additional schema information, add the schema property prefixed with @ to the comment section above the field.

Currently, following schema properties are supported:

• description
• minimum
• maximum
• exclusiveMinimum
• exclusiveMaximum
• minLength
• maxLength
• minItems
• maxItems
• example

Additionally, special data types are inferred and annotated as well:

• Int: { type: 'integer', format: 'int32' }
• BigInt: { type: 'integer', format: 'int64' }
• Float: { type: 'number', format: 'float' }
• Decimal: { type: 'number', format: 'double' }
• DateTime: { type: 'string', format: 'date-time' }

Example

This example using @description and @minimum tags

/// @description Number of reviews
/// @minimum 9
reviewCount Int?     @default(0)

will generate @ApiProperty() decorator with description and minimum as properties as well as type and format to specify the data type. If a default value is specified, it gets also added to the decorator.

@ApiProperty({
  description: 'Number of reviews',
  minimum: 9,
  type: 'integer',
  format: 'int32',
  default: 0
})
reviewCount: number | null;

Validation decorators

If classValidation = "true", the generator will add validation decorators from class-validator to each field of CreateDTO and UpdateDTO that can then be used in combination with the NestJS ValidationPipe (see NestJS Auto-validation).

Some decorators will be inferred from the field's attributes. If the field is optional, it will add @IsOptional(), otherwise @IsNotEmpty(). If the field is a list, it will add @IsArray(). Type validators are inferred from the field's type:

• String → @IsString()
• Boolean → @IsBoolean()
• Int → @IsInt()
• BigInt → @IsInt()
• Float: → @IsNumber()
• Decimal: → @IsNumber()
• DateTime → @IsRFC3339()
• Json → @IsJSON()

All remaining validation decorators can be added in the comment/documentation section above the field. The parentheses can be omitted if not passing a value.

Example

Prisma Schema

/// @Contains('Product')
name        String   @db.VarChar(255)
reviewCount Int?     @default(0)
/// @ArrayNotEmpty
tags        String[]

Generated output

@IsNotEmpty()
@IsString()
@Contains('Product')
name: string;
@IsOptional()
@IsInt()
reviewCount?: number;
@IsNotEmpty()
@IsArray()
@ArrayNotEmpty()
tags: string[];

Example

Prisma Schema

generator nestjsDto {
provider = "prisma-types-generator"
output = "../src"
outputToNestJsResourceStructure = "true"
}

model Question {
id String @id @default(dbgenerated("gen_random_uuid()")) @db.Uuid
/// @DtoReadOnly
createdAt DateTime @default(now())
/// @DtoRelationRequired
createdBy User? @relation("CreatedQuestions", fields: [createdById], references: [id])
createdById String? @db.Uuid
updatedAt DateTime @updatedAt
/// @DtoRelationRequired
updatedBy User? @relation("UpdatedQuestions", fields: [updatedById], references: [id])
updatedById String? @db.Uuid

  /// @DtoRelationRequired
  /// @DtoRelationCanConnectOnCreate
  category   Category? @relation(fields: [categoryId], references: [id])
  categoryId String?   @db.Uuid

  /// @DtoCreateOptional
  /// @DtoRelationCanCreateOnCreate
  /// @DtoRelationCanConnectOnCreate
  /// @DtoRelationCanCreateOnUpdate
  /// @DtoRelationCanConnectOnUpdate
  tags Tag[]

  title     String
  content   String
  responses Response[]

}

Generated results

// src/question/dto/connect-question.dto.ts
export class ConnectQuestionDto {
  id: string;
}

// src/question/dto/create-question.dto.ts
import { ApiExtraModels } from '@nestjs/swagger';
import { ConnectCategoryDto } from '../../category/dto/connect-category.dto';
import { CreateTagDto } from '../../tag/dto/create-tag.dto';
import { ConnectTagDto } from '../../tag/dto/connect-tag.dto';

export class CreateQuestionCategoryRelationInputDto {
  connect: ConnectCategoryDto;
}
export class CreateQuestionTagsRelationInputDto {
  create?: CreateTagDto[];
  connect?: ConnectTagDto[];
}

@ApiExtraModels(
  ConnectCategoryDto,
  CreateQuestionCategoryRelationInputDto,
  CreateTagDto,
  ConnectTagDto,
  CreateQuestionTagsRelationInputDto,
)
export class CreateQuestionDto {
  category: CreateQuestionCategoryRelationInputDto;
  tags?: CreateQuestionTagsRelationInputDto;
  title: string;
  content: string;
}

// src/question/dto/update-question.dto.ts
import { ApiExtraModels } from '@nestjs/swagger';
import { CreateTagDto } from '../../tag/dto/create-tag.dto';
import { ConnectTagDto } from '../../tag/dto/connect-tag.dto';

export class UpdateQuestionTagsRelationInputDto {
  create?: CreateTagDto[];
  connect?: ConnectTagDto[];
}

@ApiExtraModels(CreateTagDto, ConnectTagDto, UpdateQuestionTagsRelationInputDto)
export class UpdateQuestionDto {
  tags?: UpdateQuestionTagsRelationInputDto;
  title?: string;
  content?: string;
}

// src/question/entities/question.entity.ts
import { User } from '../../user/entities/user.entity';
import { Category } from '../../category/entities/category.entity';
import { Tag } from '../../tag/entities/tag.entity';
import { Response } from '../../response/entities/response.entity';

export class Question {
  id: string;
  createdAt: Date;
  createdBy?: User;
  createdById: string;
  updatedAt: Date;
  updatedBy?: User;
  updatedById: string;
  category?: Category;
  categoryId: string;
  tags?: Tag[];
  title: string;
  content: string;
  responses?: Response[];
}

Principles

Generally we read field properties from the DMMF.Field information provided by @prisma/generator-helper. Since a few scenarios don't become quite clear from that, we also check for additional annotations (or decorators) in a field's documentation (that is anything provided as a tripple slash comments for that field in your prisma.schema).

Initially, we wanted DTO classes to implement Prisma.<ModelName><(Create|Update)>Input but that turned out to conflict with required relation fields.

ConnectDTO

This kind of DTO represents the structure of input-data to expect from 'outside' (e.g. REST API consumer) when attempting to connect to a model through a relation field.

A Models ConnectDTO class is composed from a unique'd list of isId and isUnique scalar fields. If the ConnectDTO class has exactly one property, the property is marked as required. If there are more than one properties, all properties are optional (since setting a single one of them is already sufficient for a unique query) - you must however specify at least one property.

ConnectDTOs are used for relation fields in CreateDTOs and UpdateDTOs.

CreateDTO

This kind of DTO represents the structure of input-data to expect from 'outside' (e.g. REST API consumer) when attempting to create a new instance of a Model. Typically the requirements for database schema differ from what we want to allow users to do. As an example (and this is the opinion represented in this generator), we don't think that relation scalar fields should be exposed to users for create, update, or delete activities (btw. TypeScript types generated in PrismaClient exclude these fields as well). If however, your schema defines a required relation, creating an entity of that Model would become quite difficult without the relation data. In some cases you can derive information regarding related instances from context (e.g. HTTP path on the rest endpoint /api/post/:postid/comment to create a Comment with relation to a Post). For all other cases, we have the

• @DtoRelationCanCreateOnCreate
• @DtoRelationCanConnectOnCreate
• @DtoRelationCanCreateOnUpdate
• @DtoRelationCanConnectOnUpdate

annotations that generate corresponding input properties on CreateDTO and UpdateDTO (optional or required - depending on the nature of the relation).

When generating a Models CreateDTO class, field that meet any of the following conditions are omitted (order matters):

• isReadOnly OR is annotated with @DtoReadOnly (Note: this apparently includes relation scalar fields)
• field represents a relation (field.kind === 'object') and is not annotated with @DtoRelationCanCreateOnCreate or @DtoRelationCanConnectOnCreate
• field is a relation scalar
• field is not annotated with @DtoCreateOptional AND
  • isId && hasDefaultValue (id fields are not supposed to be provided by the user)
  • isUpdatedAt (Prisma will inject value)
  • isRequired && hasDefaultValue (for schema-required fields that fallback to a default value when empty. Think: createdAt timestamps with @default(now()) (see now()))

UpdateDTO

When generating a Models UpdateDTO class, field that meet any of the following conditions are omitted (order matters):

• field is annotated with @DtoUpdateOptional
• isReadOnly OR is annotated with @DtoReadOnly (Note: this apparently includes relation scalar fields)
• isId (id fields are not supposed to be updated by the user)
• field represents a relation (field.kind === 'object') and is not annotated with @DtoRelationCanCreateOnUpdate or @DtoRelationCanConnectOnUpdate
• field is a relation scalar
• field is not annotated with @DtoUpdateOptional AND
  • isId (id fields are not supposed to be updated by the user)
  • isUpdatedAt (Prisma will inject value)
  • isRequired && hasDefaultValue (for schema-required fields that fallback to a default value when empty. Think: createdAt timestamps with @default(now()) (see now()))

Entity

When generating a Models Entity class, only fields annotated with @DtoEntityHidden are omitted. All other fields are only manipulated regarding their isRequired and isNullable flags.

By default, every scalar field in an entity is required meaning it doesn't get the TypeScript "optional member flag" ? next to it's name. Fields that are marked as optional in PrismaSchema are treated as nullable - meaning their TypeScript type is a union of field.type and null (e.g. string | null).

Relation and relation scalar fields are treated differently. If you don't specifically include a relation in your query, those fields will not exist in the response.

• every relation field is always optional (isRequired = false)
• relations are nullable except when
  • the relation field is a one-to-many or many-to-many (i.e. list) type (would return empty array if no related records found)
  • the relation was originally flagged as required (isRequired = true)
  • the relation field is annotated with @DtoRelationRequired (do this when you mark a relation as optional in PrismaSchema because you don't want (SQL) ON DELETE CASCADE behavior - but your logical data schema sees this relation as required)
• default values are not added to the @ApiProperty() decorator

DTO

The plain DTO class is almost the same as Entity with the difference that all relation fields are omitted. This is useful if your response is the plain entity without any (optional) relations.