Testing GraphQL Schemas

GraphQL still has a steep learning curve when it comes to implementing an API. Not to mention, how does one go about testing each part of a GraphQL API? Since GraphQL is easily mocked due to the decoupling of resolvers from their schema, it's not all that complex to setup thorough unit tests.

Before the code though, it's important to setup some guard-rails beyond unit tests when it comes to your schema. Specifically since designing a schema (or any API) should be an iterative process. Some great tools that can aid in incrementally building a well structured Graph include:

1. A solid Linter to make sure the schema itself is properly written.
2. GraphQL Inspector to ensure that changes to the schema don't break the API's consumers.

With the those quick wins out of the way, let's focus on those aforementioned unit tests.

There is a need to mock the return data of your schema, we need to pull in a few packages from the graphql-tools library to perform this mocking. Namely the GraphQlFileLoader class from @graphql-tools/graphql-file-loader, addMockToSchema from @graphql-tools/mock, and loadSchemaSync from @graphql-tools/load These packages will allow the tests to introspect a .graphql file, mock the return values, and allow the suite's assertion matchers to work their magic. It's totally OK to install these as dev dependencies since they are not a requirement (usually) for runtime.

import { dirname, join } from 'path';
import { fileURLToPath } from 'url';
import { loadSchemaSync } from '@graphql-tools/load';
import { GraphQLFileLoader } from '@graphql-tools/graphql-file-loader';
import { addMocksToSchema } from '@graphql-tools/mock';
import { graphql } from 'graphql';

const __dirname = dirname(fileURLToPath(import.meta.url));
const schema = loadSchemaSync(join(__dirname, '../schema.graphql'), {
  loaders: [
    new GraphQLFileLoader(),
  ]
});

const mocks = {
  String: () => "Funzones",
  Boolean: () => false
}

const mockedSchema = addMocksToSchema({ schema, mocks })

With the schema loaded, the only thing left is to write a query and execute it with the graphql package.

const query = `
  query funzoneFinder {
    funzoneById(id: "512f5428-63d1-4bf4-8c23-467ac944e58d") { content { description, title } }
  }
`
const result = await graphql(mockedSchema, query)
t.deepLooseEqual(result.data, { productById: {
  content: {
    description: "Funzones",
    title: "Funzones"
  }
}}, "Funzones are actually fun")

The above matcher example utilizes the lightweight and fabulous Tape testing library. If most of an application is running outside of a browser, it's a much more lightweight and performant way of testing over Jest (IMHO)

With this setup, a team can specifically test any access pattern that their consumers might utilize. As another win, these tests don't rely on slow resolvers, meaning they execute quickly. This should lead to fast feedback as a graph is constructed.