Objects, Enums, Unions, and Interfaces

These four are common building blocks for APIs. Let's take a look at how fuse makes them easy to use for you.

Objects

In the nodes we talked about keyed entities and that we can load them by means of their key. This doesn't cover all the cases, you won't want to expose a node interface/... for every given entity in your graph and some of them might not have unique keys.

In comes the objectType, let's look at an example where our User entity has an Address associated with it.

import { addObjectFields, objectType } from 'fuse'
 
const Address = objectType<Address>({
  name: 'Address',
  fields: (t) => ({
    street: t.exposeString('name'),
    houseNumber: t.exposeInt('houseNumber'),
    city: t.exposeString('city'),
    country: t.exposeString('country'),
  }),
})
 
addObjectFields(UserNode, t => ({
  address: t.field({
    type: Address,
    resolve: (parent) => {
      // fetch address for parent.id (the user id)
    }
  })
}))

The Address has no primary key, but is contextual to the user and we'll fetch it in the context of our parent user.

You can still integrate dataloader here by switching out the t.field with t.loadable or t.loadableList if the user would have multiple addresses.

import { addNodeFields } from 'fuse'
 
addNodeFields(UserNode, (t) => ({
  address: t.loadable({
    type: Address,
    load: ids => {
      /** Load the addresses for all given user-ids */
    }
    resolve: (parent, args) => {
      /** Return the contextual user-id */
      return parent.id
    }
  }),
}))

Enums

There are types where you want to narrow down the possible values for a given field, this can be done by means of the enumType. In this case let's give our BlogPost a state, this can be published or draft:

import { enumType, addNodeFields } from 'fuse'
 
const PostStatus = enumType({
  name: 'PostStatus',
  // The "as" const is important so TypeScript knows
  // this array can't change and can easily give you
  // type-hints.
  values: ['PUBLISHED', 'DRAFT', 'UNKNOWN'] as const,
})
 
addNodeFields(BlogPostnode, t => ({
  status: t.field({
    type: PostStatus,
    resolve: (parent) => {
      if (parent.published) {
        return 'PUBLISHED'
      } else if (parent.draft) {
        return 'DRAFT'
      }
 
      return 'UNKNOWN'
    }
  })
}))

Now when the types are generated on the front-end it will know that post.status can have these three given values.

Interfaces

Our API also supports inheritance, with interfaces we can define a common shape across objects. Think about the built-in Node interface which dictates that any node implements the id property.

We can go further and define interfaces ourself as well, let's look at an example

import { node, interfaceType } from 'fuse'
 
const ContentNode = interfaceType({
  name: 'ContentNode',
  fields: (t) => ({
    title: t.string()
  }),
})
 
const BlogPostNode = node<{ id: string; title: string; content: string }>({
  name: 'BlogPost',
  interfaces: [ContentNode],
  load: async (ids) => [],
  isTypeOf: (parent: any) => {
    return !!parent.content
  },
  fields: (t) => ({
    title: t.exposeString('title'),
    content: t.exposeString('content'),
  }),
})
 
addQueryFields(t => ({
  content: t.field({
    type: [ContentNode],
    resolve: () => []
  })
}))

In the above example we define a ContentNode that tells us that every implemeentor needs to have a title property of type string, then we go on to create our content entry-point and tell us that any returned value here can be an implementor of the ContentNode type. On the BlogPost you'll see a isTypeOf function, this is needed to tell which implementor of the interface is being returned.

We can query this by doing

query {
  content {
    title
    ... on BlogPost {
      id content
    }
    __typename
  }
}

Checking the __typename on the front-end will narrow down the type and you can add logic based on the specific concrete type.

Unions

Some endpoints will return multiple possible types, with unions you can catch this case. Let's look at a case where we got an entry-point named content, this can return blogposts or advertisements

On the Content type you can see a resolveType function, this is an alternative to the aforementioned isTypeOf function from the interfaceType section.

import { node, unionType } from 'fuse'
 
const BlogPostNode = node<{ id: string, content: string }>({
  name: 'BlogPost',
  load: async (ids) => [],
  fields: (t) => ({
    content: t.exposeString('content'),
  }),
})
 
const AdvertisementNode = node<{ id: string, title: string }>({
  name: 'Advertisement',
  load: async (ids) => [],
  fields: (t) => ({
    title: t.exposeString('title'),
  }),
})
 
const Content = unionType({
  name: 'Content',
  types: [BlogPostNode, AdvertisementNode],
  // This is needed so we can tell the field what type
  // we are dealing with.
  resolveType(blogOrAdvertisement) {
    if (blogOrAdvertisement.title) {
      return 'Advertisement'
    } else {
      return 'BlogPost'
    }
  },
})
 
addQueryFields(t => ({
  content: t.field({
    type: [Content],
    resolve: () => []
  })
}))

We can query this by doing

query {
  content {
    ... on BlogPost {
      id content
    }
    ... on Advertisement {
      id title
    }
    __typename
  }
}

Checking the __typename on the front-end will narrow down the type and you can add logic based on the specific concrete type.

Handling lists Context