This example demonstrates how to ensure idempotence in a GraphQL service with the Serverless Framework, targeting AWS. It is based on the example of GraphQL API with DynamoDB.

The side effect of the function is writing the nickname. AWS DynamoDB provides an idempotent transactional write API, which writes name for only once and does nothing on retry. It needs an argument clientRequestToken to check whether the current transactional write is a retry. The clientRequestToken should be a universally unique identifier. Then we use awsRequestId to be the clientRequestToken. This is a unique identifier provided by AWS Lambda. When Lambda retries a function, it will use the same awsRequestId as that in the first invocation.

Start by initializing a project and installing the graphql module.

$ npm init
$ npm install --save graphql

Now we can use it in handler.js, where we declare a schema and then use it to serve query requests.

/* handler.js */
const {
  graphql,
  GraphQLSchema,
  GraphQLObjectType,
  GraphQLString,
  GraphQLNonNull
} = require('graphql')

// This method just inserts the user's first name into the greeting message.
const getGreeting = firstName => `Hello, ${firstName}.`

// Here we declare the schema and resolvers for the query
const schema = new GraphQLSchema({
  query: new GraphQLObjectType({
    name: 'RootQueryType', // an arbitrary name
    fields: {
      // the query has a field called 'greeting'
      greeting: {
        // we need to know the user's name to greet them
        args: { firstName: { name: 'firstName', type: new GraphQLNonNull(GraphQLString) } },
        // the greeting message is a string
        type: GraphQLString,
        // resolve to a greeting message
        resolve: (parent, args) => getGreeting(args.firstName)
      }
    }
  }),
})

// We want to make a GET request with ?query=<graphql query>
// The event properties are specific to AWS. Other providers will differ.
module.exports.query = (event, context, callback) => graphql(schema, event.queryStringParameters.query)
  .then(
    result => callback(null, {statusCode: 200, body: JSON.stringify(result)}),
    err => callback(err)
  )

Pretty simple! To deploy it, define a service in serverless.yml, and set the handler to service HTTP requests.

# serverless.yml
service: graphql-api

functions:
  query:
    handler: handler.query
    events:
      - http:
          path: query
          method: get

Now we can bring it to life:

$ serverless deploy
# Serverless: Packaging service...
# Serverless: Excluding development dependencies...
# Serverless: Uploading CloudFormation file to S3...
# Serverless: Uploading artifacts...
# Serverless: Uploading service .zip file to S3 (357.34 KB)...
# Serverless: Validating template...
# Serverless: Updating Stack...
# Serverless: Checking Stack update progress...
# ..............
# Serverless: Stack update finished...
# Service Information
# service: graphql-api
# stage: dev
# region: us-east-1
# stack: graphql-api-dev
# api keys:
#   None
# endpoints:
#   GET - https://9qdmq5nvql.execute-api.us-east-1.amazonaws.com/dev/query
# functions:
#   query: graphql-api-dev-query

$ curl -G 'https://9qdmq5nvql.execute-api.us-east-1.amazonaws.com/dev/query' --data-urlencode 'query={greeting(firstName: "Jeremy")}'
# {"data":{"greeting":"Hello, Jeremy."}}

In the real world, virtually any service that does something valuable has a data store behind it. For example, suppose users have nicknames that should appear in the greeting message. We need a database to store the nicknames, and we can expand our GraphQL API to update them.

Let's start by adding a database to the resource definitions in serverless.yml. We need a table keyed on the user's first name, which we define using CloudFormation, as well as some provider configuration to allow our function to access it.

# add to serverless.yml

provider:
  name: aws
  runtime: nodejs6.10
  environment:
    DYNAMODB_TABLE: ${self:service}-${self:provider.stage}
  iamRoleStatements:
    - Effect: Allow
      Action:
        - dynamodb:GetItem
        - dynamodb:UpdateItem
      Resource: "arn:aws:dynamodb:${opt:region, self:provider.region}:*:table/${self:provider.environment.DYNAMODB_TABLE}"

resources:
  Resources:
    NicknamesTable:
      Type: 'AWS::DynamoDB::Table'
      Properties:
        AttributeDefinitions:
          - AttributeName: firstName
            AttributeType: S
        KeySchema:
          - AttributeName: firstName
            KeyType: HASH
        ProvisionedThroughput:
          ReadCapacityUnits: 1
          WriteCapacityUnits: 1
        TableName: ${self:provider.environment.DYNAMODB_TABLE}

We need to run serverless deploy again to update the changes made in serverless.yml:

$ serverless deploy

To use it we need the aws-sdk, In this example, I use the SDK's vanilla DocumentClient to access DynamoDB records.

$ npm install --save aws-sdk

Include these in our handler, and then we can get to work.

// add to handler.js
const AWS = require('aws-sdk');
const dynamoDb = new AWS.DynamoDB.DocumentClient();

Before, we defined a method that just returned a string value for the greeting message. However, the GraphQL library can also use Promises as resolvers. Since the DocumentClient uses a callback pattern, we'll wrap these in promises and use the DynamoDB get method to check the database for a nickname for the user.

// add to handler.js
const promisify = foo => new Promise((resolve, reject) => {
  foo((error, result) => {
    if(error) {
      reject(error)
    } else {
      resolve(result)
    }
  })
})

// replace previous implementation of getGreeting
const getGreeting = firstName => promisify(callback =>
  dynamoDb.get({
    TableName: process.env.DYNAMODB_TABLE,
    Key: { firstName },
  }, callback))
  .then(result => {
    if(!result.Item) {
      return firstName
    }
    return result.Item.nickname
  })
  .then(name => `Hello, ${name}.`)

  // add method for updates
const changeNickname = (firstName, nickname) => promisify(callback =>
  dynamoDb.update({
    TableName: process.env.DYNAMODB_TABLE,
    Key: { firstName },
    UpdateExpression: 'SET nickname = :nickname',
    ExpressionAttributeValues: {
      ':nickname': nickname
    }
  }, callback))
  .then(() => nickname)

You can see here that we added a method changeNickname, but the GraphQL API is not yet using it. We need to declare a mutation that the front-end can use to perform updates. We previously only added a query declaration to the schema. Now we need a mutation as well.

// alter schema
const schema = new GraphQLSchema({
  query: new GraphQLObjectType({
    /* unchanged */
  }),
  mutation: new GraphQLObjectType({
    name: 'RootMutationType', // an arbitrary name
    fields: {
      changeNickname: {
        args: {
          // we need the user's first name as well as a preferred nickname
          firstName: { name: 'firstName', type: new GraphQLNonNull(GraphQLString) },
          nickname: { name: 'nickname', type: new GraphQLNonNull(GraphQLString) }
        },
        type: GraphQLString,
        // update the nickname
        resolve: (parent, args) => changeNickname(args.firstName, args.nickname)
      }
    }
  })
})

After these changes, we can make the greeting request again and receive the same result as before.

$ curl -G 'https://9qdmq5nvql.execute-api.us-east-1.amazonaws.com/dev/query' --data-urlencode 'query={greeting(firstName: "Jeremy")}'
# {"data":{"greeting":"Hello, Jeremy."}}

But if I want the API to call me "Jer", I can update the nickname for "Jeremy".

$ curl -G 'https://9qdmq5nvql.execute-api.us-east-1.amazonaws.com/dev/query' --data-urlencode 'query=mutation {changeNickname(firstName:
 \"Jeremy\", nickname: \"Jer\")}'
$ curl -G 'https://9qdmq5nvql.execute-api.us-east-1.amazonaws.com/dev/query' --data-urlencode 'query={greeting(firstName: \"Jeremy\")}'
# {"data":{"greeting":"Hello, Jer."}}

The API will now call anyone named "Jeremy" by the nickname "Jer". This kind of separation of concerns lets you build front-ends and services that offload logic into back-ends that use abstract data access and processing behind one, strongly typed, validated, uniform contract that comes with rich versioning and deprecation strategies.

To deploy this service yourself, download the source code and deploy it with the Serverless Framework. Happy building!