Skip to content

ExpediaGroup/graphql-component

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

GraphQL schema components.

This project is designed to facilitate componentized or modularized development of GraphQL schemas.

Read more about the idea here.

graphql-component lets you build a schema progressively through a tree (facilitated through imports) of GraphQLComponent instances. Each GraphQLComponent instance encapsulates an executable GraphQL schema, specifically a graphql-js GraphQLSchema object. See the API below, but the encapsulated schema is accessible through a simple schema getter on a given GraphQLComponent instance.

Generally speaking, each instance of GraphQLComponent has reference to an instance of GraphQLSchema. This instance of GraphQLSchema is built in a several ways, depending on the options passed to a given GraphQLComponent's constructor.

  • when a GraphQLComponent instance has imports (ie. other GraphQLComponent instances or component configuration objects) graphql-tools stitchSchemas() is used to create a "gateway" or aggregate schema that is the combination of the underlying imported schemas, and the typeDefs/resolvers passed to the root or importing GraphQLComponent
  • when a GraphQLComponent has no imports, graphql-tools' makeExecuteableSchema({typeDefs, resolvers}) is used to generate an executable GraphQL schema using the passed/required inputs.

It's worth noting that GraphQLComponent can also be used to construct componentized Apollo Federated schemas. That is, if you pass the federation: true flag to a GraphQLComponent constructor, @apollo/federation's buildSubgraphSchema() is used in lieu of graphql-tools makeExecutableSchema({...}) and the above still schema construction rule applies. The general use case here might be to help modularize an individual federated subschema service implementation.

Running the examples

local schema composition:

  • can be run with npm run start-composition

federation (2 subschema services implemented via GraphQLComponent and a vanilla Apollo Gateway):

  • can be run with npm run start-federation

Repository structure

  • lib - the graphql-component code.
  • examples/composition - a simple example of composition using graphql-component
  • examples/federation - a simple example of building a federated schema using graphql-component

Running examples:

  • composition: npm run start-composition
  • federation: npm run start-federation
  • go to localhost:4000/graphql
    • for composition this will bring up the GraphQL Playground for a plain old Apollo Server
    • for the federation example this will bring up the GraphQL Playground for an Apollo Federated Gateway

Debug output

GraphQLComponent uses debug for local stdout based debug logging. Enable all debug logging with the node environment variable DEBUG=graphql-component:*. Generally speaking, most debug output occurs during GraphQLComponent construction.

API

  • GraphQLComponent(options) - the component class, which may also be extended. Its options include:

    • types - a string or array of strings of GraphQL SDL defining the type definitions for this component
    • resolvers - a resolver map (ie. a two level map whose first level keys are types from the SDL, mapped to objects, whose keys are fields on those types and values are resolver functions)
    • imports - an optional array of imported components for the schema to be merged with.
    • context - an optional object { namespace, factory } for contributing to context.
    • directives - an optional object containing custom schema directives.
    • mocks - a boolean (to enable default mocks) or an object to pass in custom mocks
    • dataSources - an array of data sources instances to make available on context.dataSources .
    • dataSourceOverrides - overrides for data sources in the component tree.
    • federation - make this component's schema an Apollo Federated schema (default: false).
    • pruneSchema - (optional) prune the schema according to pruneSchema in graphql-tools (default: false)
    • pruneSchemaOptions - (optional) schema options as per PruneSchemaOptions in graphql-tools
  • static GraphQLComponent.delegateToComponent(component, options) - a wrapper function that utilizes graphql-tools delegateToSchema() to delegate the calling resolver's selection set to a root type field (Query, Mutuation) of another GraphQLComponent's schema

    • component (instance of GraphQLComponent) - the component's whose schema will be the target of the delegated operation

    • options (object)

      • operation (optional, can be inferred from info): query or mutation
      • fieldName (optional, can be inferred if target field has same name as calling resolver's field): the target root type (Query, Mutation) field in the target GraphQLComponent's schema
      • context (required) - the context object from resolver that calls delegateToComponent
      • info (required) - the info object from the resolver that calls delegateToComponent
      • args (object, optional) - an object literal whose keys/values are passed as args to the delegatee's target field resolver. By default, the resolver's args from which delegateToComponent is called will be passed if the target field has an argument of the same name. Otherwise, arguments passed via the args object will override the calling resolver's args of the same name.
      • transforms (optional Array<Transform>): Transform being a valid graphql-tools transform
    • please see graphql-tools delegateToSchema documentation for more details on available options since the delegateToComponent functions is simply an adapter for the GraphQLComponent API.

A GraphQLComponent instance (ie, new GraphQLComponent({...})) has the following API:

  • schema - getter that this component's GraphQLSchema object (ie. the "executable" schema that is constructed as described above)
  • context - context function that builds context for all components in the tree.
  • types - this component's types.
  • resolvers - this component's resolvers.
  • imports - this component's imported components in the form of import configuration objects
  • mocks - custom mocks for this component.
  • directives - this component's directives.
  • dataSources - this component's data source(s), if any.

General usage

Creating a component using the GraphQLComponent class:

const GraphQLComponent = require('graphql-component');

const { schema, context } = new GraphQLComponent({ types, resolvers });

Encapsulating state

Typically the best way to make a re-useable component with instance data will be to extend GraphQLComponent.

const GraphQLComponent = require('graphql-component');
const resolvers = require('./resolvers');
const types = require('./types');
const mocks = require('./mocks');

class PropertyComponent extends GraphQLComponent {
  constructor({ types, resolvers }) {
    super({ types, resolvers });
  }
}

module.exports = PropertyComponent;

Aggregation

Example to merge multiple components:

const { schema, context } = new GraphQLComponent({
  imports: [
    new Property(),
    new Reviews()
  ]
});

const server = new ApolloServer({
  schema,
  context
});

Import configuration

Imports can be a configuration object supplying the following properties:

  • component - the component instance to import.
  • exclude - fields on types to exclude from the component being imported, if any.

Exclude

You can exclude whole types or individual fields on types.

const { schema, context } = new GraphQLComponent({
  imports: [
    {
      component: new Property(),
      exclude: ['Mutation.*']
    },
    {
      component: new Reviews(),
      exclude: ['Mutation.*']
    }
  ]
});

The excluded types will not appear in the aggregate or gateway schema exposed by the root component, but are still present in the schema encapsulated by the underlying component. This can keep from leaking unintended API surface area, if desired. You can still delegate calls to imported component's schema to utilize the excluded field under the covers.

Data Source support

Data sources in graphql-component do not extend apollo-datasource's DataSource class.

Instead, data sources in components will be injected into the context, but wrapped in a proxy such that the global context will be injected as the first argument of any function implemented in a data source class.

This allows there to exist one instance of a data source for caching or other statefulness (like circuit breakers), while still ensuring that a data source will have the current context.

For example, a data source should be implemented like:

class PropertyDataSource {
  async getPropertyById(context, id) {
    //do some work...
  }
}

This data source would be executed without passing the context manually:

const resolvers = {
  Query: {
    property(_, { id }, { dataSources }) {
      return dataSources.PropertyDataSource.getPropertyById(id);
    }
  }
}

Setting up a component to use a data source might look like:

new GraphQLComponent({
  //...
  dataSources: [new PropertyDataSource()]
})

Override data sources

Since data sources are added to the context based on the constructor name, it is possible to simply override data sources by passing the same class name or overriding the constructor name:

const { schema, context } = new GraphQLComponent({
  imports: [
    {
      component: new Property(),
      exclude: ['Mutation.*']
    },
    {
      component: new Reviews(),
      exclude: ['Mutation.*']
    }
  ],
  dataSourceOverrides: [
    new class PropertyMock {
      static get name() {
        return 'PropertyDataSource';
      }
      //...etc
    }
  ]
});

Decorating the global context

Example context argument:

const context = {
  namespace: 'myNamespace',
  factory: function ({ req }) {
    return 'my value';
  }
};

After this, resolver context will contain { ..., myNamespace: 'my value' }.

Context middleware

It may be necessary to transform the context before invoking component context.

const { schema, context } = new GraphQLComponent({types, resolvers, context});

context.use('transformRawRequest', ({ request }) => {
  return { req: request.raw.req };
});

Using context now in apollo-server-hapi for example, will transform the context to one similar to default apollo-server.