Skip to content

Latest commit

 

History

History

type-merging-interfaces

Folders and files

NameName
Last commit message
Last commit date

parent directory

..
 
 
 
 
 
 
 
 
 
 

Cross-service interfaces

Watch the chapter video

Cross-service interfaces

This example explores setting up a GraphQL interface that spans across service boundaries, as described in the merged interfaces documentation. This is an easily overlooked feature made possible by the flexibility of type merging.

This example demonstrates:

  • Distributing a GraphQL interface across services.

Setup

cd cross-service-interfaces

yarn install
yarn start

The following service is available for interactive queries:

For simplicity, all subservices in this example are run locally by the gateway server. You could easily break out any subservice into a standalone remote server following the combining local and remote schemas example.

Summary

Visit the stitched gateway and try running the following query:

query {
  storefront(id: "1") {
    id
    name
    productOfferings {
      __typename
      id
      name
      price
      ...on ProductDeal {
        products {
          name
          price
        }
      }
    }
  }
}

If you study the results here, Storefront.productOfferings field returns the ProductOffering interface:

# Products schema
interface ProductOffering {
  id: ID!
  name: String!
  price: Float!
}

This interface is implemented by two types:

  • Product: a basic product record from the Products service.
  • ProductDeal: a wrapper for a set of products given a special price, managed in the Storefronts service.

The oddity here is that the Storefronts service knows Product only with an id field, therefore Storefronts is not able to implement the full interface, it may only implement a subset of it:

# Storefronts schema
interface ProductOffering {
  id: ID!
}

This is where the flexibility of type merging really shines. By virtue of the merge, ProductDeal will adopt the full ProductOffering interface in the combined gateway schema.

However, that means the gateway schema provides an interface of fields for a type that the underlying subservice does not provide. This difference in interface fields is automatically translated using typed fragments. For example, try the following query:

query {
  storefront(id: "1") {
    productOfferings {
      id
      name
      price
    }
  }
}

The above requests the full ProductOffering interface from the gateway schema. Comparing this to the query delegated to the Storefronts subservice, you'll see that the interface selections have been expanded into typed fragments that are compatible with the Storefronts schema:

query ($graphqlTools0__v0_id: ID!) {
  graphqlTools0_storefront: storefront(id: $graphqlTools0__v0_id) {
    productOfferings {
      id
      ... on Product {
        __typename
        id
      }
      ... on ProductDeal {
        __typename
        name
        price
      }
      __typename
    }
  }
}