Graphlient

A friendlier Ruby client for consuming GraphQL-based APIs. Built on top of your usual graphql-client, but with better defaults, more consistent error handling, and using the faraday HTTP client.

Table of Contents

• Installation
• Usage
  • Schema Storing and Loading on Disk
  • Preloading Schema Once
  • Error Handling
  • Executing Parameterized Queries and Mutations
  • Parse and Execute Queries Separately
  • Dynamic vs. Static Queries
  • Generate Queries with Graphlient::Query
  • Create API Client Classes with Graphlient::Extension::Query
  • Swapping the HTTP Stack
  • Testing with Graphlient and RSpec
• License

Installation

Add the following line to your Gemfile.

gem 'graphlient'

Usage

Create a new instance of Graphlient::Client with a URL and optional headers/http_options.

client = Graphlient::Client.new('https://test-graphql.biz/graphql',
  headers: {
    'Authorization' => 'Bearer 123'
  },
  http_options: {
    read_timeout: 20,
    write_timeout: 30
  }
)

http_options	default	type
read_timeout	nil	seconds
write_timeout	nil	seconds

The schema is available automatically via .schema.

client.schema # GraphQL::Schema

Make queries with query, which takes a String or a block for the query definition.

With a String.

response = client.query <<~GRAPHQL
  query {
    invoice(id: 10) {
      id
      total
      line_items {
        price
        item_type
      }
    }
  }
GRAPHQL

With a block.

response = client.query do
  query do
    invoice(id: 10) do
      id
      total
      line_items do
        price
        item_type
      end
    end
  end
end

This will call the endpoint setup in the configuration with POST, the Authorization header and query as follows.

query {
  invoice(id: 10) {
    id
    total
    line_items {
      price
      item_type
    }
  }
}

A successful response object always contains data which can be iterated upon. The following example returns the first line item's price.

response.data.invoice.line_items.first.price

You can also execute mutations the same way.

response = client.query do
  mutation do
    createInvoice(input: { fee_in_cents: 12_345 }) do
      id
      fee_in_cents
    end
  end
end

The successful response contains data in response.data. The following example returns the newly created invoice's ID.

response.data.create_invoice.first.id

Schema storing and loading on disk

To reduce requests to graphql API you can cache schema:

client = Client.new(url, schema_path: 'config/your_graphql_schema.json')
client.schema.dump! # you only need to call this when graphql schema changes

Preloading Schema Once

Even if caching the schema on disk, instantiating Graphlient::Client often can be both time and memory intensive due to loading the schema for each instance. This is especially true if the schema is a large file. To get around these performance issues, instantiate your schema once and pass it in as a configuration option.

One time in an initializer

schema = Graphlient::Schema.new(
  'https://graphql.foo.com/graphql', 'lib/graphql_schema_foo.json'
)

Pass in each time you initialize a client

client = Graphlient::Client.new(
  'https://graphql.foo.com/graphql',
  schema: schema,
  headers: {
    'Authorization' => 'Bearer 123',
  }
)

Error Handling

Unlike graphql-client, Graphlient will always raise an exception unless the query has succeeded.

• Graphlient::Errors::ClientError: all client-side query validation failures based on current schema
• Graphlient::Errors::GraphQLError: all GraphQL API errors, with a humanly readable collection of problems
• Graphlient::Errors::ExecutionError: all GraphQL execution errors, with a humanly readable collection of problems
• Graphlient::Errors::ServerError: all transport errors raised by HTTP Adapters. You can access inner_exception, status_code and response on these errors to get more details on what went wrong
• Graphlient::Errors::FaradayServerError: this inherits from ServerError ☝️, we recommend using ServerError to rescue these
• Graphlient::Errors::HttpServerError: this inherits from ServerError ☝️, we recommend using ServerError to rescue these
• Graphlient::Errors::ConnectionFailedError: this inherits from ServerError ☝️, we recommend using ServerError to rescue these
• Graphlient::Errors::TimeoutError: this inherits from ServerError ☝️, we recommend using ServerError to rescue these
• Graphlient::Errors::HttpOptionsError: all NoMethodError raised by HTTP Adapters when given options in http_options are invalid

All errors inherit from Graphlient::Errors::Error if you need to handle them in bulk.

Executing Parameterized Queries and Mutations

Graphlient can execute parameterized queries and mutations by providing variables as query parameters.

The following query accepts an array of IDs.

With a String.

query = <<-GRAPHQL
  query($ids: [Int]) {
    invoices(ids: $ids) {
      id
      fee_in_cents
    }
  }
GRAPHQL
variables = { ids: [42] }

client.query(query, variables)

With a block.

client.query(ids: [42]) do
  query(ids: [:int]) do
    invoices(ids: :ids) do
      id
      fee_in_cents
    end
  end
end

Graphlient supports following Scalar types for parameterized queries by default:

• :id maps to ID
• :boolean maps to Boolean
• :float maps to Float
• :int maps to Int
• :string maps to String

You can use any of the above types with ! to make it required or use them in [] for array parameters.

For any other custom types, graphlient will simply use to_s of the symbol provided for the type, so query(ids: [:InvoiceType!]) will result in query($ids: [InvoiceType!]).

The following mutation accepts a custom type that requires fee_in_cents.

client.query(input: { fee_in_cents: 12_345 }) do
  mutation(input: :createInvoiceInput!) do
    createInvoice(input: :input) do
      id
      fee_in_cents
    end
  end
end

Parse and Execute Queries Separately

You can parse and execute queries separately with optional variables. This is highly recommended as parsing a query and validating a query on every request adds performance overhead. Parsing queries early allows validation errors to be discovered before request time and avoids many potential security issues.

# parse a query, returns a GraphQL::Client::OperationDefinition
query = client.parse do
  query(ids: [:int]) do
    invoices(ids: :ids) do
      id
      fee_in_cents
    end
  end
end

# execute a query, returns a GraphQL::Client::Response
client.execute query, ids: [42]

Or pass in a string instead of a block:

# parse a query, returns a GraphQL::Client::OperationDefinition
query = client.parse <<~GRAPHQL
  query($some_id: Int) {
    invoice(id: $some_id) {
      id
      feeInCents
    }
  }
GRAPHQL

# execute a query, returns a GraphQL::Client::Response
client.execute query, ids: [42]

Dynamic vs. Static Queries

Graphlient uses graphql-client, which recommends building queries as static module members along with dynamic variables during execution. This can be accomplished with graphlient the same way.

Create a new instance of Graphlient::Client with a URL and optional headers.

module SWAPI
  Client = Graphlient::Client.new('https://test-graphql.biz/graphql',
    headers: {
      'Authorization' => 'Bearer 123'
    },
    allow_dynamic_queries: false
  )
end

The schema is available automatically via .schema.

SWAPI::Client.schema # GraphQL::Schema

Define a query.

module SWAPI
  InvoiceQuery = Client.parse do
    query(id: :int) do
      invoice(id: :id) do
        id
        fee_in_cents
      end
    end
  end
end

Execute the query.

response = SWAPI::Client.execute(SWAPI::InvoiceQuery, id: 42)

Note that in the example above the client is created with allow_dynamic_queries: false (only allow static queries), while graphlient defaults to allow_dynamic_queries: true (allow dynamic queries). This option is marked deprecated, but we're proposing to remove it and default it to true in graphql-client#128.

Generate Queries with Graphlient::Query

You can directly use Graphlient::Query to generate raw GraphQL queries.

query = Graphlient::Query.new do
  query do
    invoice(id: 10) do
      line_items
    end
  end
end

query.to_s
# "\nquery {\n  invoice(id: 10){\n    line_items\n    }\n  }\n"

Use of Fragments

Fragments should be referred by constant:

module Fragments
  Invoice = client.parse <<~'GRAPHQL'
    fragment on Invoice {
      id
      feeInCents
    }
  GRAPHQL
end

Graphlient offers the syntax below to refer to the original constant:

• Triple underscore ___ to refer to the fragment
• Double underscore __ for namespace separator

In this example, Fragments::Invoice would be referred as follows:

invoice_query = client.parse do
  query do
    invoice(id: 10) do
      id
      ___Fragments__Invoice
    end
  end
end

The wrapped response only allows access to fields that have been explicitly asked for. In this example, while id has been referenced directly in the main query, feeInCents has been spread via fragment and trying to access it in the original wrapped response will throw GraphQL::Client::ImplicitlyFetchedFieldError (to prevent data leaks between components).

response = client.execute(invoice_query)
result = response.data.invoice
result.to_h
# {"id" => 10, "feeInCents"=> 20000}
result.id
# 10
result.fee_in_cents
# raises GraphQL::Client::ImplicitlyFetchedFieldError

feeInCents cannot be fetched directly from the main query, but from the fragment as shown below:

invoice = Fragments::Invoice.new(result)
invoice.id
# 10
invoice.fee_in_cents
# 20000

Create API Client Classes with Graphlient::Extension::Query

You can include Graphlient::Extensions::Query in your class. This will add a new method_missing method to your context which will be used to generate GraphQL queries.

include Graphlient::Extensions::Query

query = query do
  invoice(id: 10) do
    line_items
  end
end

query.to_s
# "\nquery{\n  invoice(id: 10){\n    line_items\n    }\n  }\n"

Swapping the HTTP Stack

You can swap the default Faraday adapter for Net::HTTP.

client = Graphlient::Client.new('https://test-graphql.biz/graphql',
  http: Graphlient::Adapters::HTTP::HTTPAdapter
)

Testing with Graphlient and RSpec

Use Graphlient inside your RSpec tests in a Rails application or with Rack::Test against your actual application.

require 'spec_helper'

describe App do
  include Rack::Test::Methods

  def app
    # ...
  end

  let(:client) do
    Graphlient::Client.new('http://test-graphql.biz/graphql') do |client|
      client.http do |h|
        h.connection do |c|
          c.adapter Faraday::Adapter::Rack, app
        end
      end
    end
  end

  context 'an invoice' do
    let(:result) do
      client.query do
        query do
          invoice(id: 10) do
            id
          end
        end
      end
    end

    it 'can be retrieved' do
      expect(result.data.invoice.id).to eq 10
    end
  end
end

Alternately you can stub_request with Webmock.

describe App do
  let(:url) { 'http://example.com/graphql' }
  let(:client) { Graphlient::Client.new(url) }

  before do
    stub_request(:post, url).to_return(
      status: 200,
      body: DummySchema.execute(GraphQL::Introspection::INTROSPECTION_QUERY).to_json
    )
  end

  it 'retrieves schema' do
    expect(client.schema).to be_a Graphlient::Schema
  end
end

In order to stub the response to actual queries, dump the schema into a JSON file and specify it via schema_path as follows.

describe App do
  let(:url) { 'http://graph.biz/graphql' }
  let(:client) { Graphlient::Client.new(url, schema_path: 'spec/support/fixtures/invoice_api.json') }
  let(:query) do
    <<~GRAPHQL
      query{
        invoice(id: 42) {
          id
          feeInCents
        }
      }
    GRAPHQL
  end
  let(:json_response) do
    {
      'data' => {
        'invoice' => {
          'id' => '42',
          'feeInCents' => 2000
        }
      }
    }.to_json
  end

  before do
    stub_request(:post, url).to_return(
      status: 200,
      body: json_response
    )
  end

  it 'returns invoice fees' do
    response = client.query(query)
    expect(response.data).to be_truthy
    expect(response.data.invoice.id).to eq('42')
    expect(response.data.invoice.fee_in_cents).to eq(2000)
  end
end

License

MIT License, see LICENSE