graphql prepare should generate binding initializer

[kbrandwijk]

Currently, graphql prepare generates the static binding class, but it doesn't generate the initialization part of it (e.g. new Binding(...)). This means you have to manually duplicate the endpoint, headers and schema location in code.

To improve this workflow, the graphql prepare command could generate this for you.

Now the naive implementation would take whatever graphql-config comes up with, and use that. However, there are two catches here.

First, graphql-config will resolve a prisma endpoint through the prisma extension, so it will have an actual endpoint URL and authorization header. Code generation however, should use prisma-binding, not graphql-binding, because of additional features for Prisma.

Second, graphql-config resolves environment variables. However, if someone is using environment variables in their .graphqlconfig, it's highly likely that they would also want to use env vars in the code.

The implementation proposal (that will follow) should take these two points into account.

[kbrandwijk]

First, the basic use case without env vars:

So let's take the following .graphqlconfig configuration:

projects:
  app:
    schemaPath: src/schema.graphql
    extensions:
      endpoints:
        default: http://localhost:4000
  database:
    schemaPath: src/generated/prisma.graphql
    extensions:
      prisma: database/prisma.yml
      prepare-binding:
        output: src/generated/prisma.ts
        generator: prisma-ts
  bahnhof:
    schemaPath: src/generated/bahnhof.graphql
    extensions:
      endpoints:
        default:
          url: 'https://api.deutschebahn.com/free1bahnql/v1/graphql'
          headers:
            Authorization: Bearer myToken
      prepare-binding:
        output: src/generated/bahnhof.ts
        generator: binding-ts

To use these bindings in my code, I will need the the following for bahnhof (see also dotansimha/graphql-binding#30):

const link = new HttpLink({ uri, fetch })
  const middlewareLink = (operation, forward) => {
    console.log('operation', operation)
    if (authorizationToken) {
      operation.setContext(context => ({
        ...context,
        headers: {
          ...context.headers,
          Authorization: `Bearer ${authorizationToken}`
        }
      }))
    }
    return forward(operation)
  }

  const executableSchema = makeRemoteExecutableSchema({
    schema: fs.readFileSync('src/generated/bahnhof.graphql', 'utf-8'),
    link: ApolloLink.from([middlewareLink, link]),
  });

const binding = new Binding({ schema: executableSchema })

And the following for database:

const binding = new Prisma({
      endpoint: 'https://eu1.prisma.sh/public-patternraccoon-961/my-service/dev',
      secret: 'mysecret123', // specified in database/prisma.yml
      debug: true, // log all GraphQL queries & mutations
    }),

Now if graphql prepare could create a code file with init() methods, the server file would look like:

import { initDatabaseBinding, initBahnhofBinding } from 'src/generated/initBindings.ts'

const server = new GraphQLServer({
  typeDefs: './src/schema.graphql',
  resolvers,
  context: req => ({
    ...req,
    db: initDatabaseBinding(),
    bahnhof: initBahnhofBinding()
  }),
})

[Urigo]

We've just released a new version and an alpha release channel for GraphQL CLI - 4.0.0-alpha.XXX!

In order to receive schemas the new new GraphQL Config which let's you specify multiple schemas from any source possible.

I want to ask the people on this issue, in that case, why would you need the prepare command?
Are we missing anything and should we add this command back?

Checkout new instructions and the migration guide and we will be happy to hear your feedback during the alpha period!

Feel free to contact us by opening an issue or using our Discord channel;
https://discord.gg/xud7bH9

[Urigo]

I've posted the wrong Discord link, this is the correct one: https://discord.gg/xud7bH9

[wtrocki]

This command is no longer supported in 4.0.0. Closing