graphql-typescript-definitions

Generate TypeScript definition files from .graphql documents.

Installation

yarn add graphql-typescript-definitions

Usage

This package will generate matching .d.ts files for each .graphql file you specify. It will generate types in the following format:

• A default export for the type that will be generated by a GraphQL loader (GraphQL’s DocumentNode type, but augmented as graphql-typed’s DocumentNode which includes additional type details about the operation).

• An interface for each query, mutation, and fragment, named Data. For example, query Home {} becomes export interface HomeQueryData {}.

• A namespace for each operation that includes any nested types. Nested types are named in pascal case using their keypath from the root of the operation. For example, if we imagine the following GraphQL schema (using the GraphQL IDL):

  type Person {
    name: String!
    relatives: [Person!]!
  }
  
  type Query {
    person: Person
  }
  

  and the following query:

  query Someone {
    person {
      name
      relatives {
        name
      }
    }
  }
  

  The following exports would be generated:

  export interface SomeoneQueryData {
    person?: SomeoneQueryData.Person | null;
  }
  
  export namespace SomeoneQueryData {
    export interface Person {
      name: string;
      relatives: SomeoneQueryData.PersonRelatives[];
    }
  
    export interface PersonRelatives {
      name: string;
    }
  }
  

  This allows you to use the full query’s type, as well as any of the subtypes that make up that query type. This is particularly useful for list or nullable types, where you can directly the access the underlying type without any additional help from TypeScript:

  import someoneQueryDocument, {SomeoneQueryData} from './Someone.graphql';
  
  let data: SomeoneQueryData;
  let person: SomeoneQueryData.Person;
  

Operation

On startup this tool performs the following actions:

• Loads all schemas
• Extracts all enums, input objects, and custom scalars as schema types
• Writes the schema types to types.ts (or ${projectName}-types.ts for named projects)
  • Written in directory provided by --schema-types-path argument
  • Override --schema-types-path per project with the schemaTypesPath extension

Configuration

This tool reads schema information from a graphql-config config file in the project root.

Examples

A project configuration with a schemaTypesPath override

{
  "schema": "build/schema.json",
  "include": ["app/**/*.graphql"],
  "extensions": {
    "schemaTypesPath": "app/bar/types/graphql"
  }
}

Type Generation

Nullability

As demonstrated in the root person field in the example above, nullable fields are represented as optional types, in a union with null. Nullable items in list fields (i.e., [Person]!) are represented as a union type with null.

Interfaces and Unions

Interface an union fields are represented as union types in cases where there are spreads that could result in different fields on different concrete types. The type names for these cases are named the same as the default naming (pascal case version of the keypath for the field), but with the type condition appended to the end. All cases not covered by fragments are extracted into a type with a postpended Other name.

# Schema
interface Named {
  name: String!
}

type Person implements Named {
  name: String!
  occupation: String
}

type Dog implements Named {
  name: String!
  legs: Int!
}

type Cat implements Named {
  name: String!
  livesLeft: Int!
}

type Horse implements Named {
  name: String!
  topSpeed: Float!
}

type Query {
  named: Named
}

# Query
query SomeNamed {
  named {
    name
    ... on Person {
      occupation
    }
    ... on Dog {
      legs
    }
  }
}

// generated types
export interface SomeNamedData {
  named?:
    | SomeNamedData.NamedPerson
    | SomeNamedData.NamedDog
    | SomeNamedData.NamedOther
    | null;
}

export namespace SomeNamedData {
  export interface NamedPerson {
    __typename: 'Person';
    name: string;
    occupation?: string | null;
  }
  export interface NamedDog {
    __typename: 'Dog';
    name: string;
    legs: number;
  }
  export interface NamedOther {
    __typename: 'Cat' | 'Horse';
    name: string;
  }
}

Note that the above example assumes that you specify the --add-typename argument. These types are only useful when a typename is included either explicitly or with this argument, as otherwise there is no simple way for TypeScript to disambiguate the union type.

Schema Types

Input types (enums, input objects, and custom scalars) are generated once, in a central location, and imported within each typing file. You can use these definitions to reference the schema types in other application code as well. If --enum-style=enum (the default) is set then GraphQL enums are turned into corresponding TypeScript enums, else if --enum-style=type is set then GraphQL enums are created as string union types. The schema types directory is specified using the --schema-types-path argument (detailed below), and the format for the generated typescript enums can be specified using the --enum-format option.

CLI

graphql-typescript-definitions --schema-types-path app/types

As noted above, the configuration of your schema and GraphQL documents is done via a .graphqlconfig file, as this allows configuration to shared between tools. The CLI does support a few additional options, though:

• --schema-types-path: specifies a directory to write schema types (REQUIRED)
• --watch: watches the include globbing patterns for changes and re-processes files (default = false)
• --cwd: run tool for .graphqlconfig located in this directory (default = process.cwd())
• --add-typename: adds a __typename field to every object type (default = true)
• --export-format: species the shape of values exported from .graphql files (default = document)
  • Options: document (exports a graphql-typed DocumentNode), typedDocumentNode (exports a @graphql-typed-document-node/core TypedDocumentNode), documentWithTypedDocumentNode (exports a union of a graphql-typed DocumentNode and a @graphql-typed-document-node/core TypedDocumentNode),simple (exports a graphql-typed SimpleDocument),
• --enum-format: specifies output format for enum types (default = undefined)
  • Options: camel-case, pascal-case, snake-case, screaming-snake-case
  • undefined results in using the unchanged name from the schema (verbatim)
• --enum-style: specifies if the graphql enum types should be created as typescript enums or string unions types. (default = 'enum')
  • Options: enum (exports typescript enums), type (exports a string union type, that has no runtime impact)
• --custom-scalars: specifies custom types to use in place of scalar types in your GraphQL schema. See below for details.

Examples

# run tool for .graphqlconfig in current directory, produces ./app/graphql/types
graphql-typescript-definitions --schema-types-path app/graphql/types

# run watcher for .graphqlconfig in current directory, produces ./app/graphql/types
graphql-typescript-definitions --schema-types-path app/graphql/types --watch

# run tool for .graphqlconfig in a child directory, produces ./src/app/graphql/types
graphql-typescript-definitions --cwd src --schema-types-path app/graphql/types

--custom-scalars

By default, all custom scalars are exported as an alias for string. You can export a different type for these scalars by passing in a --custom-scalars option. This option is a JSON-serialized object that specifies what custom type to import from a package and re-export as the type for that scalar. For example, assuming the following schema:

scalar HtmlString

You may want a custom TypeScript type for any field of this GraphQL type (for example, to restrict functions to use only this type, and not any arbitrary string). Assuming you have an installed npm package by the name of my-custom-type-package, and this package exports a named SafeString type, you could pass the following --custom-scalars option:

yarn run graphql-typescript-definitions --schema-path 'build/schema.json' --schema-types-path 'src/schema' --custom-scalars '{"HtmlString": {"name": "SafeString", package: "my-custom-type-package"}}'

With this configuration, your custom scalar would be exported roughly as follows:

import {SafeString} from 'my-custom-type-package';
export type HtmlString = SafeString;

You can also use built-in types by simply leaving off the package property:

yarn run graphql-typescript-definitions --schema-path 'build/schema.json' --schema-types-path 'src/schema' --custom-scalars '{"Seconds": {"name": "number"}}'

This will produce a simple type alias:

export type Seconds = number;

Node

const {Builder} = require('graphql-typescript-definitions');

const builder = new Builder({
  schemaTypesPath: 'app/graphql/types',
});

builder.on('build', (build) => {
  // See the source file for details on the shape of the object returned here
  console.log(build);
});

builder.on('error', (error) => {
  console.error(error);
});

// Optionally, you can pass {watch: true} here to re-run on changes
builder.run();

As with the CLI, you can pass options to customize the build and behavior:

• watch
• enumFormat (use the exported EnumFormat enum)
• enumStyle (use the exported EnumStyle enum)
• graphQLFiles
• schemaPath
• schemaTypesPath
• customScalars
• config (custom GraphQLConfig instance)

Customizing filesystem interactions

In projects with thousands of GraphQL documents, Builder may take a few seconds to reach its ready state. Additionally, very large projects with deep file trees may exceed Node's default memory limits (especially in non-MacOS environments). To allow project/environment-specific customization, builders accept a GraphQLFilesystem object.

const {
  AbstractGraphQLFilesystem,
  Builder,
} = require('graphql-typescript-definitions');

class CustomFilesystem extends AbstractGraphQLFilesystem {
  watch(config) {
    // Set up filesystem watchers, and emit change:schema / change:document / delete:document events.
  }

  getGraphQLProjectIncludedFilePaths(projectConfig) {
    // Find all .graphql documents for the given project.
  }

  dispose() {
    // Release any resources held by watchers.
  }
}

const builder = new Builder({
  graphQLFilesystem: new CustomFilesystem(),
});

builder.run({watch: true});