Skip to Content
Yoga
DocumentationFeaturesCustom Plugins

Plugins

Yoga builds upon the GraphQL plugin system envelop and adds additional HTTP specific hooks for customizing both the GraphQL execution and HTTP layer. Many of the additional features of Yoga are added via plugins. If you need to customize your GraphQL API you can also write your own plugins.

Using Plugins

You can both use Yoga or Envelop plugins with GraphQL Yoga.

💡

When there is both a Envelop and Yoga specific plugin available, you should always opt-in for the Yoga variant as this one allows certain optimizations compared to just an Envelop plugin. As envelop plugins have only control over the GraphQL execution for example a plugin such as the response-cache , can never act on the HTTP level. However, the yoga response-cache, Persisted Operations or Defer/Stream plugin utilizes the HTTP level hooks, and thus can completly skip all of the expensive GraphQL execution steps for serving cached results.

Yoga Plugin Example

Please refer to the corresponding feature documentation pages. E.g. Response Cache, Persisted Operations or Defer/Stream.

Yoga Defer Stream Plugin Usage Example
import { createServer } from 'node:http'
import { createYoga } from 'graphql-yoga'
import { useDeferStream } from '@graphql-yoga/plugin-defer-stream'
 
const yoga = createYoga({
  plugins: [useDeferStream()]
})
 
const server = createServer(yoga)
 
server.listen(4000, () => {
  console.info('Server is running on http://localhost:4000/graphql')
})

Envelop Plugin Example

💡

Envelop is a lightweight JavaScript (TypeScript) library for customizing the GraphQL execution layer, allowing developers to build, share and compose plugins that enhance the capabilities of your GraphQL server . GraphQL Yoga uses Envelop  under the hood so you can easily extend your server’s capabilities with the plugins from Envelop Ecosystem 

The following example adds GraphQL JIT  to our GraphQL Server using Envelop GraphQL JIT Plugin 

import { createSchema, createYoga } from 'graphql-yoga'
import { useGraphQlJit } from '@envelop/graphql-jit'
 
// Provide your schema
const yoga = createYoga({
  schema: createSchema({
    typeDefs: /* GraphQL */ `
      type Query {
        greetings: String!
      }
    `,
    resolvers: {
      Query: {
        greetings: () => 'Hello World!'
      }
    }
  }),
  plugins: [useGraphQlJit()]
})
 
// Start the server and explore http://localhost:4000/graphql
const server = createServer(yoga)
server.listen(4000, () => {
  console.info('Server is running on http://localhost:4000/graphql')
})

Writing Plugins

Sometimes you might want to build your own plugins. You can write your own Yoga plugin and even share it with other people by publishing it to npm.

💡

A good entry-point for discovering how to write Yoga plugins is to look at the source code of the existing plugins maintained by us.

The most hooks for Yoga origin from the envelop plugin system. Please refer to the Envelop Plugin Lifecycle documentation for more information.  In addition, Yoga adds more HTTP specific hooks.

Plugin Lifecycle

The following diagram shows the plugin lifecycle of GraphQL Yoga. For a detailed description of each hook, please refer to the detail sections of each hook.

onRequest

This hook is invoked for ANY incoming HTTP request. Here you can manipulate the request or create a short circuit before Yoga handles the request.

⚠️

Exceptions thrown by this hook are not caught. This means they will buble up to the HTTP server underlying implementation.

For example, the node:http server crashes the entire process on uncaught exceptions.

Prefer onRequestParse when possible, or wrap the hook code in a try block.

Example actions in this hook:

  • Manipulate the request
  • Short circuit before Yoga handles the request

API

Example

import { Plugin } from 'graphql-yoga'
 
function useAuth(): Plugin {
  return {
    onRequest({ request, fetchAPI, endResponse }) {
      if (!request.headers.get('authorization')) {
        endResponse(
          new fetchAPI.Response(null, {
            status: 401,
            headers: {
              'Content-Type': 'application/json'
            }
          })
        )
      }
    }
  }
}

onRequestParse

This hook is invoked for any incoming GraphQL HTTP request and is invoked before attempting to parse the GraphQL parameters. Here you can manipulate the request, set a custom request parser or apply security measures such as checking for access tokens etc.

Example actions in this hook:

  • Manipulate the request
  • Set custom GraphQL request parser (use custom GraphQL protocol)
  • Apply security measures

onParams

This hook is invoked for an incoming GraphQL request after the GraphQL parameters (query, variables, extensions and operationName) have been ATTEMPTED to be parsed.

Within this hook you can manipulate and customize the parameters or even implement a whole new way of parsing the parameters (if you wish to diverge from the GraphQL over HTTP specification).

In addition to that you could also short-circuit and skip the GraphQL execution. E.g. you might want to serve a result from the cache instead.

Example actions in this hook:

  • Response Cache: Short-circuit GraphQL execution if response can be served from the cache.
  • Persisted Operations: Load the query document string from the persisted operations store before running the execution.
  • APQ: Load/persist the query document string on the persisted operations store.

onParse

Called for parsing the GraphQL document. This hook has a before and after stage. You can hook into before the document is parsed and/or into after the document is parsed.

Example actions in this hook:

  • Replace GraphQL parser and support future GraphQL syntax
  • Collect metrics about parsing success/failures

onValidate

Called for validating the GraphQL document. This hook has a before and after stage. You can hook into before the document is parsed and/or into after the document is parsed.

Example actions in this hook:

  • Register additional validation rules (e.g. Disable Introspection)
  • Collect metrics about validation success/failures

onContextBuilding

Called for building the GraphQL context. This hook has a before and after stage. You can hook into before and after the context is built.

Example actions in this hook:

  • Authentication
  • Add data to context object (e.g. user data or dataloader instances)

onExecute

Called for executing a GraphQL mutation or query operation. This hook has a before and after stage. You can hook into before and after the GraphQL request is executed.

Example actions in this hook:

  • Collect metrics about execution time
  • Error logging/reporting

onSubscribe

Called for subscribing to a GraphQL subscription operation. This hook has a before and after stage. You can hook into before and after the GraphQL subscription is executed.

Example actions in this hook:

  • Collect metrics about execution time
  • Error logging/reporting

onExecutionResult

This hook is invoked for each result produced for GraphQL operation, before it is processed to be sent to client.

In particular, it is useful to handle batched operations. If a request contains batched operations, this hook is called once of each operation, while onResultProcess will be only called once for the entire request.

Here, you can modify the result, to add monitoring or instrumentation extensions for example.

Example actions in this hook:

  • Add metadata to results
  • Collect errors

onResultProcess

This hook is invoked after a GraphQL request has been processed and before the response is forwarded to the client. Here you can customize what transport/response processor format should be used for sending the result over the wire.

Example actions in this hook:

  • Specify custom response format
  • Logging/Metrics

onResponse

This hook is invoked after a HTTP request (both GraphQL and NON GraphQL) has been processed and after the response has been forwarded to the client. Here you can perform any cleanup or logging operations, or you can manipulate the outgoing response object.

⚠️

Exceptions thrown by this hook are not caught. This means they will buble up to the HTTP server underlying implementation.

For example, the node:http server crashes the entire process on uncaught exceptions.

Prefer onRequestParse when possible, or wrap the hook code in a try block.

Example actions in this hook:

  • Specify custom response format
  • Logging/Metrics

API

Example

import { Plugin } from 'graphql-yoga'
 
function useYogaSignature(): Plugin {
  return {
    onResponse({ request, response }) {
      response.headers.set('X-GraphQL-Server', 'Yoga')
    }
  }
}

onDispose

If you have plugins that need some internal resources to be disposed of, you can use the onDispose hook to dispose of them. This hook will be invoked when the GraphQL Yoga instance is disposed like above.

let dbConnection: Connection
const plugin = {
  onPluginInit: async () => {
    dbConnection = await createConnection()
  },
  onDispose: async () => {
    // Dispose of resources
    await dbConnection.close()
  }
}

Or you can flush a queue of operations when the server is shutting down.

const backgroundJobs: Promise<void>[] = []
 
const plugin = {
  onRequestParse() {
    backgroundJobs.push(
      sendAnalytics({
        /* ... */
      })
    )
  },
  onDispose: async () => {
    // Flush the queue of background jobs
    await Promise.all(backgroundJobs)
  }
}

But for this kind of purposes, waitUntil can be a better choice. Learn more about explicit resource management here.

instrumentation

An optional instrumentation instance can be present in the plugin.

This Instrumentation instance allows to wrap an entire phase execution (including all plugin hooks), meaning running code just before, just after and around the execution of the phase.

Instrumentation doesn’t have access to input/output of a phase, use hooks to have access to those data. If needed, we recommend to share data between instrumentation and hooks with a WeakMap and the given context as the key.

All instrumentation takes 2 parameters:

  • payload: an object containing the graphql context or the http request depending on the instrument.
  • wrapped: The function representing the execution of the phase. It takes no parameters, and returns void (or Promise<void> for asynchrone phases). This function must always be called. If this function returns a Promise, the instrument should return a Promise resolving after it.

Instrumentation composition

If multiple plugins have instrumentation, they are composed in the same order they are defined the plugin array (the first is outtermost call, the last is inner most call).

It is possible to customize this composition if it doesn’t suite your need (ie. you need hooks and instrumentation to have a different oreder of execution).

import { composeInstrumentation, envelop } from '@envelop/core'
 
const { instrumentation: instrumentation1, ...plugin1 } = usePlugin1()
const { instrumentation: instrumentation2, ...plugin2 } = usePlugin2()
 
const instrumentation = composeInstrumentation([instrumentation2, instrumentation1])
 
const getEnveloped = envelop({
  plugin: [{ insturments }, plugin1, plugin2]
})

request

Wraps the HTTP request handling. This includes all the plugins onRequest and onResponse hooks.

This instrument can be asynchronous, the wrapped funcion can be asynchronous. Be sure to return a Promise if wrapped() returned a Promise.

requestParse

Wraps the parsing of the request phase to extract grapqhl params. This include all the plugins onRequestParse hooks.

This insturment can be asynchronous, the wrapped function can be asynchrounous. Be sure to return a Promise if wrapped() returns a Promise.

operation

Wraps the Graphql operation execution pipeline. This is called for each graphql operation, meaning it can be called mutliple time for the same HTTP request if batching is enabled.

This instrument can be asynchronous, the wrapped function can be asynchronous. Be sur to return a Promise if wrapped() returnd a Promise.

init

Wraps the envelop (the call to envelop function) initialisation.

This includes all the plugins onEnveloped hooks, and the creation of the Envelop Orchestrator.

This instrumentation must be synchrone, the wrapped function is always synchrone.

parse

Wraps the parse phase. This includes all the plugins onParse hooks and the actual engine parse.

This instrument must be synchrone, the wrapped function is always synchrone.

validate

Wraps the validate phase. This includes all the plugins onValidate hooks and the actual engine validate.

This instrument must be synchrone, the wrapped function is always synchrone.

context

Wraps the context building phase. This includes all the plugins onContextBuilding hooks.

This instrument must be synchrone, the wrapped function is always synchrone.

execute

Wraps the execute phase. This includes all the plugins onExecute hooks.

This instrument can be asynchrone, the wrapped function can be asynchrone. Be sure to await or use .then on the result of the wrapped function to run code after the execute phase.

Note that wrapped is not guaranted to return a promise.

subscribe

Wraps the subscribe phase. This includes all the plugins onSubsribe hooks. Note that it doesn’t wrap the entire lifetime of the subscription, but only it’s intialisation.

This instrument can be asynchrone, the wrapped function can be asynchrone. Be sure to await or use .then on the result of the wrapped function to run code after the subsribe phase.

Note that wrapped is not guaranted to return a promise.

resultProcess

Wraps the context result processing phase. This includes all the plugins onResultProcess hooks.

This instrumentation can be asynchrone, the wrapped function can be asynchronous. Be sure to return a Promise if wrapped() returns a Promise.

Last updated on
Explicit Resource ManagementMonitoring