What is decapi?

decapi is a set of decorators for creating GraphQL APIs in typescript. Write your types and GQL schema at once killing two birds with one stone.

• Documentation

Examples:

• Basic Express example
  • also showcases Custom decorators / Higher order decorators
• Typeorm integration example
• Forward resolution - eg. query only needed db fields
• Nested mutations or queries
• Serverless eg. AWS Lambda

Basic example

Example below is able to resolve such query

1query {
2  hello(name: "Bob") # will resolve to 'Hello, Bob!'
3}

1import { SchemaRoot, Query, compileSchema } from 'decapi'
2
3@SchemaRoot()
4class SuperSchema {
5  @Query()
6  hello(name: string): string {
7    return `Hello, ${name}!`
8  }
9}
10
11const compiledSchema = compileSchema(SuperSchema)

compiledSchema is a regular GQL executable schema compatible with graphql-js library.

To use it with apollo-server, you'd have to use like this:

1import ApolloServer from 'apollo-server'
2
3const server = new ApolloServer({ schema, graphiql: true })
4
5server.listen(3000, () =>
6  console.log('Graphql API ready on http://localhost:3000/graphql')
7)

Although it is encouraged to prefer fastify as it is a bit faster when used with jit.

Adding nested types

For now, our query field returned scalar (string). Let's return something more complex. Schema will look like:

1mutation {
2  createProduct(name: "Chair", price: 99.99) {
3    name
4    price
5    isExpensive
6  }
7}

Such query will have a bit more code and here it is:

1import {
2  SchemaRoot,
3  Query,
4  ObjectType,
5  Field,
6  Mutation,
7  compileSchema
8} from 'decapi'
9
10@ObjectType({ description: 'Simple product object type' })
11class Product {
12  @Field()
13  name: string
14
15  @Field()
16  price: number
17
18  @Field()
19  isExpensive() {
20    return this.price > 50
21  }
22}
23
24@SchemaRoot()
25class SuperSchema {
26  @Mutation()
27  createProduct(name: string, price: number): Product {
28    const product = new Product()
29    product.name = name
30    product.price = price
31    return product
32  }
33}
34
35const compiledSchema = compileSchema(SuperSchema)

Type inference from typescript

These cases are supported by decapi 2/typescript-rtti:

• Unions - for example we want to specify whether field is nullable or not
• Function returns type of Promise<SomeType>
• List (Array) type is used

All other code-first libraries on decorators like typegraphql or typegql require you to write types for these twice. Decapi infers types from typescript without any extra effort on your end.

Even in decapi 2 onward you still can write an explicit type. There are situations when typescript types are not precise enough- for example you want to be explicit about if some number type is Float or Int (GraphQLFloat or GraphQLInt).

Let's modify our Product so it has additional categories field that will return array of strings. For the sake of readability, let's ommit all fields we've defined previously.

1@ObjectType()
2class Product {
3  @Field({ type: [String] }) // note we can use any native type like GraphQLString!
4  categories(): string[] {
5    return ['Tables', 'Furniture']
6  }
7}

We've added { type: [String] } as @Field options. Type can be anything that is resolvable to GraphQL type

• Native JS scalars: String, Number, Boolean, Date.
• Any type that is already compiled to graphql eg. GraphQLFloat or any type from external graphql library etc
• Any class decorated with @ObjectType
• an enum registered with registerEnum
• Single element array for list types eg. [String] or [GraphQLFloat] or [MyEnum]

Writing Asynchronously

Every field function we write can be async and return Promise. Let's say, instead of hard-coding our categories, we want to fetch it from some external API:

1@ObjectType()
2class Product {
3  @Field({ type: [String] }) // note we can use any native type like GraphQLString!
4  async categories(): Promise<string[]> {
5    const categories = await api.fetchCategories()
6    return categories.map((cat) => cat.name)
7  }
8}

Upgrading from 1.0.0 to 2.0.0

Firstly, you must install all the new peer dependencies. This major was a complete rewrite of the reflection of types. From 2.0.0 decapi uses typescript-rtti to infer graphql types from typescript. This should work for all TS types. If you encounter a type which cannot be inferred, please raise an issue. This means you should always have your decorators without explicit type property.

Comparisons to other libraries

decapi does reflection through typescript-rtti, so it can always infer a type directly from typescript without having to write them twice. This is what sets it apart, but there are other differences:

type-graphql

There is a much more popular library with the same goals. Differences:

1. decapi does reflection through typescript-rtti, so it can always infer a type directly from typescript without having to write them twice
2. Decapi has smaller API surface-it only has hooks on top of the basic decorators for constructing schemas. Whereas type-graphql has authorization, middleware, guards.
3. Also decapi supports graphql v16. Typegraphql is still only supporting Graphql v15

typegql

• decapi has @DuplexObjectType and @DuplexField
• decapi supports interfaces and mixins
• decapi can infer Date type
• decapi casts plain objects by default, which means it is much easier to use without ORMs. Even prisma needs this casting, because all the objects returned are POJO objects.
• InputObjectType argument passed to Field/Query method is not just a plain object, but an instance of it's class.
• decapi allows you to have an empty object type-you can populate it with fields at runtime

TC39 decorators proposal

Please if you enjoy decapi, go and star repo with the proposal-decorators . That's a way to show the TC39 members that you are using decorators and you want to have them in the language.