Skip to content
Node GraphQL Conventions
Branch: develop
Clone or download
unorthodoxgeek Release/0.28.0 (#83) (#86)
* bump logging version

* GLOB-28139 - allowing multiple requireArgs sets (#82)

Why?
Some service calls are called with different args in different context. This
change allows for multiple versions of the requireArgs parameter, without
breaking current usage of the API.

What I've done:
1. instead of using a single requireArgs list, we check against an array
of requiredArgs, and if one of the versions fits the passed arguments, we cache
2. added tests for multiple versions, while making sure current tests still pass

* bump to 0.28.0
Latest commit 07b70ae Jan 14, 2019
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
.circleci releasing 0.25.0 (#70) Sep 12, 2018
.globality releasing 0.25.0 (#70) Sep 12, 2018
src GLOB-28139 - allowing multiple requireArgs sets (#82) Jan 14, 2019
.babelrc
.bumpversion.cfg Release/0.28.0 (#83) (#86) Jan 14, 2019
.eslintrc Initial adoption of nodule-config for factories. Mar 28, 2018
.gitignore Bump version and add publish config to package JSON Jun 8, 2018
.npmrc add .npmrc Mar 27, 2018
LICENSE Initial commit Mar 19, 2018
README.md Move errors/basic routes/pagination out of Nodule-GraphQL Jun 7, 2018
package.json Release/0.28.0 (#83) (#86) Jan 14, 2019
yarn.lock bump logging version Dec 28, 2018

README.md

@globality/nodule-graphql

Node GraphQL Conventions

While GraphQL APIs can be extremely versatile and expressive, there are still good reasons to define multiple such APIs, especially when API consumers have fundamentally different needs, authorization strategies, or resource definitions.

nodule-graphql aims to define conventions for building GraphQL APIs in node so that each such API can focus on its resources and resolvers instead of boilerplate and plumbing.

nodule-graphql achieves these conventions both by providing useful functions for common GraphQL operations and by automatically wiring commmon components into a bottlejs container using the nodule-config library.

Bindings

By importing nodule-graphql, the following bindings are made available.

Middleware

  • middleware.cors injects a CORS middleware with sane caching configuration for OPTIONS headers
  • middleware.helmet injects some sane security-oriented HTTP header defaults
  • middleware.requestId injects a request id into every request

Routes

  • routes.graphql returns a configured GraphQL endpoint
  • routes.graphiql returns a configured GraphiQL endpoint

Utilities

  • terminal injects a colorized terminal writing utility to help with application startup

Resources

nodule-graphql only makes ONE assumption about resource structure: the graphql.schema binding defines a reference to your overall schema:

const { GraphQLObjectType, GraphQLSchema, GraphQLString } = require('graphql');
const { bind } = require('@globality/nodule-config');

bind('graphql.schema', () => new GraphQLSchema({
    query: new GraphQLObjectType({
        name: 'HelloWorld',
        fields: {
            helloWorld: {
                type: GraphQLString,
                resolve: () => 'hello world',
            },
        },
    }),
}));

Routes

nodule-graphql provides default routes (as described above under "Bindings"), enabling very simple wiring of a GraphQL API (assuming graphql.schema is bound):

const bodyParser = require('body-parser');
const { getContainer, Nodule } = require('@globality/nodule-config');
require('@globality/nodule-graphql');

new Nodule({ name: 'hello' }).load().then(() => {
    const { express, graphql } = getContainer('routes');

    express.use(
        '/graphql',
         bodyParser.json(),
         graphql,
    );

    express.listen(8080, '127.0.0.1');
});

Resolvers

Part of the beauty of GraphQL is that resolvers are just functions. However, resolvers can be more than just function and often should be:

  • Resolvers ought to have context for tracing and debugging.

  • Resolvers should be able to perform authorization of requests.

  • Resolvers both aggregate (asynchronously) over sources of truth and translate (synchronously) these results into resources; these actions should be explicit.

Errors

Resolvers should throw errors when something fails. Within nodule-graphql, it is expected that:

  • All errors define code that can be used by API consumers for error handling business logic
  • Most errors will borrow from HTTP error codes (because they have well-known, useful semantics)
  • Error codes should be visible to API consumers via error.extensions

Services

nodule-graphql allows you to wrap your OpenAPI client implementations to add modifications to all defined endpoints.

For usage see the README in the services folder.

Local Development

Local development of nodule-graphql with other repos has a few common pitfalls related to the usage of peer dependencies:

  • nodule-config is a peer-dependency because various libraries act as plugins to it and it needs a single import of bottlejs to share plugin state

  • graphql is a peer-dependency because it validates that all graphql definitions come from the same import and will fail otherwise

To work with nodule-graphql locally:

  1. Run yarn build within nodule-graphql to transpile the source.

  2. Change directories to your local repo that you want to test against nodule-graphql.

  3. Run yarn add /path/to/nodule-graphql to copy the transpiled source into your local repo. Do NOT use yarn link

  4. After running yarn add, remove (or move-of-the-way) the nodule_modules from within nodule_modules/@globality/nodule-graphql/

You can’t perform that action at this time.