Migration from Yoga V1

Installation

You can start with updating graphql-yoga package.

yarn add graphql-yoga@three

Server setup

Yoga v1 no longer uses a class for constructing the server, the createYoga function is now used. Also the typeDefs and resolvers config options must now be passed to a schema property with createSchema.

Yoga v1

import { GraphQLServer } from 'graphql-yoga'
import { typeDefs, resolvers } from './schema'
 
const server = new GraphQLServer({ typeDefs, resolvers })
 
server.start()

Yoga v3

import { createYoga, createSchema } from 'graphql-yoga'
import { createServer } from 'http'
import { typeDefs, resolvers } from './schema'
 
const yoga = createYoga({
  schema: createSchema({ typeDefs, resolvers })
})
 
const server = createServer(yoga)
server.listen(4000, () => {
  console.info('Server is running on http://localhost:4000/graphql')
})

Load type definitions from a file

In Yoga v1 it was possible to provide a file path for the typeDefs.

Yoga v1

import * as path from 'path'
import { GraphQLServer } from 'graphql-yoga'
import { resolvers } from './schema'
 
const server = new GraphQLServer({
  typeDefs: path.join(__dirname, 'type-definitions.graphql'),
  resolvers
})
 
server.start()

In Yoga v3 you now need to use the fs module for Node.js.

Yoga v3

import * as path from 'path'
import * as fs from 'fs'
import { createYoga, createSchema } from 'graphql-yoga'
import { createServer } from 'http'
import { resolvers } from './schema'
 
const yoga = createYoga({
  schema: createSchema({
    typeDefs: fs.readFileSync(
      path.join(__dirname, 'type-definitions.graphql'),
      'utf-8'
    ),
    resolvers
  })
})
 
const server = createServer(yoga)
server.listen(4000, () => {
  console.info('Server is running on http://localhost:4000/graphql')
})

For more complex loading of type-definitions please refer to graphql-tools/load-files .

Schema directives (previously `directiveResolvers`)

In Yoga v1 you could pass a legacy graphql-tools directiveResolver implementation to the constructor.

Yoga v1

import { GraphQLServer } from 'graphql-yoga'
import { resolvers } from './schema'
import { uppercaseDirectiveResolverImplementation } from './uppercase-directive-resolver-implementation'
 
const server = new GraphQLServer({
  typeDefs: /* GraphQL */ `
    type Query {
      hi: String
    }
  `,
  directiveResolvers: uppercaseDirectiveResolverImplementation
})
 
server.start()

Before deciding upon using schema directives, you should consider whether your custom directive could be instead implemented via a field argument (abstraction).

In Yoga v3 you have to leverage the mapSchema API from graphql-tools.

yarn add @graphql-tools/utils
yarn add @graphql-tools/schema

Yoga v3

import { Plugin, createYoga } from 'graphql-yoga'
import { createServer } from 'http'
import { mapSchema, getDirective, MapperKind } from '@graphql-tools/utils'
import { makeExecutableSchema } from '@graphql-tools/schema'
import { defaultFieldResolver } from 'graphql'
import { typeDefs, resolvers } from './schema'
 
let schema = makeExecutableSchema({
  typeDefs: [
    typeDefs,
    /* GraphQL */ `
      directive @uppercase on FIELD_DEFINITION
    `
  ],
  resolvers
})
 
schema = mapSchema(schema, {
  [MapperKind.OBJECT_FIELD]: (fieldConfig) => {
    const upperDirective = getDirective(schema, fieldConfig, 'uppercase')?.[0]
    if (upperDirective) {
      const { resolve = defaultFieldResolver } = fieldConfig
      return {
        ...fieldConfig,
        resolve: async function (source, args, context, info) {
          const result = await resolve(source, args, context, info)
          if (typeof result === 'string') {
            return result.toUpperCase()
          }
          return result
        }
      }
    }
  }
})
 
const yoga = createYoga({ schema })

You can learn more about this practice within the graphql-tools schema directives documentation.

Context

In GraphQL Yoga v2 you can use the context property in the same way as GraphQL Yoga v1. The value returned from the context factory will be merged with the initial context.

The request property within the initial context is now a Fetch API Request. It can be used for accessing all the HTTP request parameters, such as headers or the method (POST, GET).

You can learn more about the context within the context documentation.

Yoga v1

import { GraphQLServer } from 'graphql-yoga'
import { typeDefs, resolvers } from './schema'
import { db } from './db'
 
const server = new GraphQLServer({
  typeDefs,
  resolvers,
  context(initialContext) {
    const authHeader = initialContext.request.headers['authorization'] ?? null
    return { ...initialContext, db, authHeader }
  }
})
 
const server = createServer(yoga)
server.listen(4000, () => {
  console.info('Server is running on http://localhost:4000/graphql')
})

Yoga v3

import { createYoga, createSchema } from 'graphql-yoga'
import { createServer } from 'http'
import { typeDefs, resolvers } from './schema'
import { db } from './db'
 
const yoga = createYoga({
  schema: createSchema({ typeDefs, resolvers }),
  context(initialContext) {
    const authHeader =
      initialContext.request.headers.get('authorization') ?? null
 
    return { db, authHeader }
  }
})
 
const server = createServer(yoga)
server.listen(4000, () => {
  console.info('Server is running on http://localhost:4000/graphql')
})

Middlewares

GraphQL Yoga v1 included graphql-middleware for wrapping resolver functions with common logic. GraphQL Yoga v2 no longer includes graphql-middleware by default as using it can result in bad performance as it wraps all field resolvers within the schema.

If you cannot migrate your graphql-middleware code to something like graphql-tool mapSchema, we recommend using the @envelop/graphql-middleware plugin.

yarn add @envelop/graphql-middleware

Yoga v1

import { GraphQLServer } from 'graphql-yoga'
import { typeDefs, resolvers } from './schema'
 
// Middleware - Permissions
 
const code = 'supersecret'
const isLoggedIn = async (resolve, parent, args, ctx, info) => {
  // Include your agent code as Authorization: <token> header.
  const permit = ctx.request.get('Authorization') === code
 
  if (!permit) {
    throw new Error(`Not authorized!`)
  }
 
  return resolve()
}
 
const permissions = {
  Query: {
    secured: isLoggedIn
  },
  Me: isLoggedIn
}
 
const server = new GraphQLServer({
  typeDefs,
  resolvers,
  middleware: [permissions]
})
 
server.start()

Yoga v3

import { createYoga, createSchema } from 'graphql-yoga'
import { createServer } from 'http'
import { useGraphQLMiddleware } from '@envelop/graphql-middleware'
import { typeDefs, resolvers } from './schema'
 
// Middleware - Permissions
 
const code = 'supersecret'
const isLoggedIn = async (resolve, parent, args, ctx, info) => {
  // Include your agent code as Authorization: <token> header.
  const permit = ctx.request.get('Authorization') === code
 
  if (!permit) {
    throw new Error(`Not authorized!`)
  }
 
  return resolve()
}
 
const permissions = {
  Query: {
    secured: isLoggedIn
  },
  Me: isLoggedIn
}
 
const yoga = createYoga({
  schema: createSchema({ typeDefs, resolvers }),
  plugins: [useGraphQLMiddleware([permissions])]
})
 
const server = createServer(yoga)
server.listen(4000, () => {
  console.info('Server is running on http://localhost:4000/graphql')
})

For more details please refer to the

Replacing GraphQL Shield

If you are using graphql-shield you might wanna have a look and see whether the following plugins might replace it:

@envelop/generic-auth: authentication and simple (directive-based) authorization on field level
@envelop/operation-field-permissions: granular field permission access based on schema coordinates
GraphQL AuthZ: a modern and flexible GraphQL authorization layer

Subscriptions

GraphQL Yoga v1 uses the old and deprecated subscriptions-transport-ws protocol. GraphQL Yoga v2+ comes with built in subscription support over SSE (Server Sent Events). One benefit of this is that you no longer need an additional library on your frontend as the SSE protocol is just simple HTTP.

Because of the protocol change you must migrate your GraphQL clients that execute GraphQL subscription operations to use the new protocol. Please use the code snippets for your GraphQL client as listed on the handle subscription on the client documentation.

Advantages of SSE over Websockets

Transported over simple HTTP instead of a custom protocol
Built in support for re-connection and event-id Simpler protocol
No trouble with corporate firewalls doing packet inspection

Advantages of Websockets over SSE

Real time, two directional communication.

SSE gotchas

Maximum open connections limit (when not using http/2)

PubSub

With GraphQL Yoga v1 used the unmaintained package graphql-subscriptions for the PubSub implementation. In GraphQL Yoga v2+, a new maintained PubSub implementation is built-in.

Yoga v1

import { PubSub } from 'graphql-yoga'
const pubSub = new PubSub()

Yoga v2

import { createPubSub } from 'graphql-yoga'
const pubSub = createPubSub()

Type-safe PubSub Usage

The old PubSub implementation was not type-safe. Now it is possible to define all the events and payloads. For a full reference please check out the Subscription PubSub documentation.

Yoga v1

import { GraphQLServer, PubSub } from 'graphql-yoga'
 
const pubSub = new PubSub()
 
const server = new GraphQLServer({
  context: { pubSub },
  typeDefs: /* GraphQL */ `
    type Query {
      _: Boolean
    }
 
    type Subscription {
      randomNumber: Int!
    }
 
    type Mutation {
      publishRandomNumber(randomNumber: Int!): Boolean
    }
  `,
  resolvers: {
    Subscription: {
      randomNumber: {
        subscribe: (_, _2, context) => {
          return context.asyncIterator('randomNumber')
        },
        resolve: (value) => value
      }
    },
    Mutation: {
      publishRandomNumber: (_, args, context) => {
        context.pubSub.publish('randomNumber', args.randomNumber)
      }
    }
  }
})
 
server.start()

Yoga v3

import { createPubSub, createYoga, createSchema } from 'graphql-yoga'
import { createServer } from 'http'
 
const pubSub = new PubSub<{
  randomNumber: [randomNumber: number]
}>()
 
const yoga = createYoga({
  context: { pubSub },
  schema: createSchema({
    typeDefs: /* GraphQL */ `
      type Query {
        _: Boolean
      }
 
      type Subscription {
        randomNumber: Int!
      }
 
      type Mutation {
        publishRandomNumber(randomNumber: Int!): Boolean
      }
    `,
    resolvers: {
      Subscription: {
        randomNumber: {
          subscribe: (_, _2, context) => {
            return context.subscribe('randomNumber')
          },
          resolve: (value) => value
        }
      },
      Mutation: {
        publishRandomNumber: (_, args, context) => {
          context.pubSub.publish('randomNumber', args.randomNumber)
        }
      }
    }
  })
})
 
const server = createServer(yoga)
server.listen(4000, () => {
  console.info('Server is running on http://localhost:4000/graphql')
})

Filtering events

Instead of the withFilter function you can now use the more modular pipe and filter functions exported from graphql-yoga. You can learn more about filtering and mapping values in the subscription filter and map values documentation.

Yoga v1

import { GraphQLServer, PubSub, withFilter } from 'graphql-yoga'
 
const pubSub = new PubSub()
 
const server = new GraphQLServer({
  context: { pubSub },
  typeDefs: /* GraphQL */ `
    type Query {
      _: Boolean
    }
 
    type Subscription {
      randomNumber(greaterThan: Int!): Int!
    }
 
    type Mutation {
      publishRandomNumber(randomNumber: Int!): Boolean
    }
  `,
  resolvers: {
    Subscription: {
      randomNumber: {
        subscribe: withFilter(
          (_, _2, context) => {
            return context.asyncIterator('randomNumber')
          },
          (payload, args) => payload > args
        ),
        resolve: (value) => value
      }
    },
    Mutation: {
      publishRandomNumber: (_, args, context) => {
        context.pubSub.publish('randomNumber', args.randomNumber)
      }
    }
  }
})
 
server.start()

Yoga v3

import { createPubSub, createYoga, pipe, filter } from 'graphql-yoga'
import { createServer } from 'http'
 
const pubSub = new PubSub<{
  randomNumber: [randomNumber: number]
}>()
 
const yoga = createYoga({
  context: { pubSub },
  schema: createSchema({
    typeDefs: /* GraphQL */ `
      type Query {
        _: Boolean
      }
 
      type Subscription {
        randomNumber(greaterThan: Int!): Int!
      }
 
      type Mutation {
        publishRandomNumber(randomNumber: Int!): Boolean
      }
    `,
    resolvers: {
      Subscription: {
        randomNumber: {
          subscribe: (_, args, context) => {
            return pipe(
              context.subscribe('randomNumber'),
              filter((value) => value > args.greaterThan)
            )
          },
          resolve: (value) => value
        }
      },
      Mutation: {
        publishRandomNumber: (_, args, context) => {
          context.pubSub.publish('randomNumber', args.randomNumber)
        }
      }
    }
  })
})
 
const server = createServer(yoga)
server.listen(4000, () => {
  console.info('Server is running on http://localhost:4000/graphql')
})

Last updated on October 6, 2022

Express GraphQL Yoga v2

Migration from Yoga V1

Installation

Server setup

Load type definitions from a file

Schema directives (previously directiveResolvers)

Context

Middlewares

Replacing GraphQL Shield

Subscriptions

Advantages of SSE over Websockets

Advantages of Websockets over SSE

SSE gotchas

PubSub

Type-safe PubSub Usage

Filtering events

Schema directives (previously `directiveResolvers`)