INTRODUCTION👋

Since Facebook released its previously internally used query language GraphQL in 2015, it has seen an outstanding increase in adoptions for all kinds of applications.

So it didn't come as a surprise when AWS built a managed solution called AppSync and released it in 2018. AppSync is becoming more and more popular and is already being used by large companies, including BMW and Sky.

GRAPHQL🎯

REST APIs can become a burden for clients with an ever growing number of endpoints. Also, there's maybe never a perfectly fitting one to only fetch the data actually needed, so clients often end up receiving more data than necessary.

GraphQL was built to solve these problems by enabling the client to specify how to structure the data when it is returned by the server.

CACHING⚡

AppSync allows you to cache data on the server-side and by that enjoying several benefits:

• In-memory caching
• Improved performance
• Decreased latency

In the background, ElastiCache for Redis is used.

Caching can be used at different levels:

• Full-request: $context.identity and $context.identity is used as key for caches
• Per-resolver: Each resolver will be cached individual keys based on its context

AMPLIFY🔺

Amplify allows you to generate VTL resolvers from your data sources automatically.

With their CLI you can simply define an annotated GraphQL schema and all of your resolvers are created automatically.

Simply start with amplify add api and run through the wizard with selecting GraphQL as your target service. You'll be asked if you want to edit your schema in your favorite text editor.

With amplify push the API will be deployed to AWS so you can immediately start using it.

DATA SOURCES & RESOLVER🔌

A resolver defines how AppSync processes your queries and mutations. It's the glue between your schema and the actual data sources.

One of the main advantages of AppSync is that you can actively integrate with different AWS services without writing any code.

A resolver is attached to a field, for example to the orders of a customer.

COGNITO INTEGRATION🔒

AppSync natively integrates with Cognito, which allows you to do fine-grained access control at a per-field level.

REAL-TIME DATA VIA SOCKETS🔌

Subscriptions give you the ability to send real-time data to your clients without the need of any additional services. It is natively integrated into AppSync.

For every mutation your client gets updated automatically. For example in React simply use useSubscription.

You can also use enhanced filtering via JSON in the response mapping template of the subscriptions resolvers. Even multiple rules can be used together either in the same filter (AND) or while having multiple filters inside a group (OR).

MONITORING📊

CloudWatch already collects a lot of metrics for you, including the number of HTTP 4xx and 5xx errors, request latency, the number of API requests and caching statistics like hits, misses or evictions.

For detailed insights about resolvers, you have to enable logging, which comes with many options.

Verbose logging for all requests will generate very deep insights, but also increase the number of logs by an order of magnitude. Be careful, as CloudWatch ingestion is pricy.

For easy observability out-of-the-box, make use of external services like Dashbird.io which doesn't require you to create custom alarms or notifications.

DATA TYPES💎

Every GraphQL object in AppSync has a name and fields. Fields in turn can have sub-fields.

Each field's type must resolve to a scalar type, either from the default set of GraphQL or from AppSync's extensions prefixed with AWS, or to one of your own custom object types.

AUTHORIZERS🔐

You can protect your API with an Authorizer. Your authorizer can be one of the type:

• API_KEY - authorized via API Key
• AWS_LAMBDA - authorized by a Lambda function
• AWS_IAM - authorized via AWS IAM
• OPENID_CONNECT - authorized via OpenID Connect
• AMAZON_COGNITO_USER_POOLS - authorized via Cognito identity pool

This secures your whole API. You can also set authorizers on resolvers (API, JWT, Lambda or Cognito) for example if you want to secure your Customer with IAM but you can add the stripe_id with aws_iam_. Similar to all other methods.

SCHEMA📝

Schema files contain your GraphQL definitions including query, mutation & subscription.

You can create your own object types which include fields with scalar types or even other object types. Meaning, you can deeply nest objects.

By adding mutations, you define how items will be inserted or updated and what are the return values.

Subscriptions allow you to listen to certain events in your API, e.g. a creation of a new customer.

CODEGEN⚙️

Codegen is a CLI tool that allows you to create types, queries, and mutations directly from your GraphQL schema.

You simply need to define where your schema is located (schema-graphql) and define where to save your generated files & it's done. ✅

For all custom types like Customer and for example you can simply define custom scalars and map them to specific types.

Codegen has tons of plugins out there for creating all sorts of types and code from your GraphQL schema.