Skip to content

annkissam/common_graphql_client

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

CommonGraphqlClient (CGC)

Module Version Hex Docs Total Download License Last Updated

An Elixir library for generating GraphQL clients.

Adapters are provided for both HTTP (using HTTPoison) and WebSockets (using AbsintheWebSocket). Both adapters support GraphQL queries, whereas WebSockets are required for subscriptions.

This library also supports client-side query validation using nodejs.

Contents

Documentation

Docs can be found at https://hexdocs.pm/common_graphql_client.

A complete walkthrough can be found on the Annkissam Alembic. It also has an associated demo.

Installation

Add :common_graphql_client to your list of dependencies in mix.exs:

def deps do
  [
    {:common_graphql_client, "~> 0.6.0"},
    {:httpoison, "~> 1.1"},            # If using HTTP queries
    {:absinthe_websocket, "~> 0.2.0"}, # If using WebSocket subscriptions (or WebSocket queries)
  ]
end

An example Mix config:

config :my_app, MyAppApi.Context,
  client: MyAppApi.Client,
  query_caller: CommonGraphQLClient.Caller.Http,             # If using HTTP queries. You can also use the WebSocket Caller.
  http_api_url: "http://127.0.0.1:4000/api",                 # The URL for the HTTP Client
  subscription_caller: CommonGraphQLClient.Caller.WebSocket, # If using WebSocket subscriptions
  websocket_api_url: "ws://127.0.0.1:4000/socket/websocket"  # The URL for the WebSocket Client

(optional) If you're using absinthe_websocket, it has a supervisor that must be added to your supervision tree. This will be application specific:

    children = [
      ...
    ] ++ [MyAppApi.Client.supervisor()]

    Supervisor.init(children, strategy: :one_for_one)

Context

The main entry point to the client will be the context. You can add additional methods, but it knows about [:list, :list_by, :get, :get_by] and their corresponding ! methods.

The context is also responsible for implementing a subscibe/0 method (if using subscriptions). That method is called when the initial connection is made (to initiate any subscriptions) and on re-connection (to re-establish the subscriptions). It can also perform any initiation that needs to happen when the connection is established (for instance, syncing missing data). After the subscription is made, each notification will call the receive\2 method.

defmodule MyAppApi.Context do
  use CommonGraphQLClient.Context,
    otp_app: :my_app

  # Identical to calling MyAppApi.Context.list(:employees)
  def list_employees do
    list(:employees)
  end

  def find_employee_by_email!(email) do
    get_by(:employees, %{employee_email: email})
  end

  def subscribe do
    # NOTE: This will call __MODULE__.receive(:employee_created, employee) when data is received
    client().subscribe_to(:employee_created, __MODULE__)

    # (optional)
    # sync_missing_data()
  end

  def receive(:employee_created, employee) do
    # do something with the created employee
  end
end

Your use case might necessitate the receive\2 method exist on another module. The second parameter of subscribe_to allows a module to be specified. The code changes would look like this:

defmodule MyAppApi.Context do
  ...

  def subscribe do
    client().subscribe_to(:employee_created, EmployeeNotificationHandler)
  end
end

defmodule EmployeeNotificationHandler do
  def receive(:employee_created, employee) do
    ...
  end
end

Client

Your application will need a client. It will be responsible for turning symbols into various GraphQL Queries and Subscriptions. It'll also map the returned results into Ecto schemas. By calling use CommonGraphQLClient.Client, several methods will be made available. The client is responsible for implementing handle\2, handle\3, and handle_subscribe_to\2 methods for each call the context makes:

defmodule MyAppApi.Client do
  use CommonGraphQLClient.Client,
    otp_app: :my_app,
    mod: MyAppApi.Context

  defp handle(:list, :employees) do
    do_post(
      :employees,
      MyAppApi.Schema.Employee,
      MyAppApi.Query.Employee.list()
    )
  end

  defp handle(:get, :employee, id),
    do: handle(:get_by, :employee, %{id: id})

  defp handle(:get_by, :employee, variables) do
    do_post(
      :employee,
      MyAppApi.Schema.Employee,
      MyAppApi.Query.Employee.get_by(variables),
      variables
    )
  end

  defp handle_subscribe_to(:employee_created, mod) do
    do_subscribe(
      mod,
      :employee_created,
      MyAppApi.Schema.Employee,
      MyAppApi.Subscription.Employee.employee_created()
    )
  end
end

Ecto Schemas

The client will map results into an ecto schema:

defmodule MyAppApi.Schema.Employee do
  use CommonGraphQLClient.Schema

  api_schema do
    field :id, :integer
    field :name, :string
    field :email, :string
  end

  @cast_params ~w(
    id
    name
    email
  )a

  def changeset(struct, attrs) do
    struct
    |> cast(attrs, @cast_params)
  end
end

By adjusting the changeset you can also map GraphQL associations into additional structs. For example, using cast_embed/3:

defmodule MyAppApi.Schema.Employee do
  use CommonGraphQLClient.Schema

  api_schema do
    field :id, :integer
    field :name, :string

    embeds_many :email_records, EmailRecord do
      field :email, :string
    end
  end

  @cast_params ~w(
    id
    name
  )a

  def changeset(struct, attrs) do
    struct
    |> cast(attrs, @cast_params)
    |> cast_embed(:email_records, with: &email_record_changeset/2)
  end

  defp email_record_changeset(struct, attrs) do
    struct
    |> cast(attrs, [:email])
  end
end

GraphQL Queries

The example client suggests organizing GraphQL Queries using modules. A module approach would look like this:

defmodule MyAppApi.Query.Employee do
  @moduledoc """
  Employee GraphQL queries
  """

  @doc false
  def list do
    """
    query {
      employees {
        id
        name
        email
      }
    }
    """
  end

  def get_by(%{email: _}) do
    """
    query get_employee($email: String) {
      employee(email: $email) {
        id
        name
      }
    }
    """
  end
end

GraphQL Subscriptions

Similar to queries, the code to initiate subscriptions can be organized using modules:

defmodule MyAppApi.Subscription.Employee do
  @moduledoc """
  Subscription adapter module Employee
  """

  @doc false
  def employee_created do
    """
    subscription {
      employee_created {
        id
        name
        email
      }
    }
    """
  end
end

Security

Client Security

The HTTP Client can send Bearer tokens, whereas the WebSocket can send a token as a query param. Since these credentials should not be in source control, this library provides a way to set them at runtime. First, update the Mix config:

config :my_app, MyAppApi.Context,
  ...
  load_from_system_env: true # add this

Second, update your client:

use CommonGraphQLClient.Client,
  ...
  http_api_token_func: fn -> System.get_env("YOUR_API_TOKEN") || raise "ENV Not Set: YOUR_API_TOKEN ENV" end
  websocket_api_token_func: fn -> System.get_env("YOUR_API_TOKEN") || raise "ENV Not Set: YOUR_API_TOKEN ENV" end

And finally, call the init\0 function from your application supervisor:

defmodule MyApp.Supervisor do
  use Supervisor

  def start_link(opts) do
    Supervisor.start_link(__MODULE__, :ok, opts)
  end

  def init(:ok) do
    MyAppApi.Client.init()

    ...
  end
end

Alternatively, http_api_token\0 and websocket_api_token\0 can be overridden in the client to support other use cases.

HTTP Server

While this package supports creating clients, if you're also building the GraphQL API in Phoenix we can make some (simple) suggestions. These examples will use a shared token. They'll also use secure_compare to mitigate timing attacks. Your API will require more complexity if it needs to support multiple users or to differentiate between clients.

A plug added to the router can secure your GraphQL API endpoint:

pipeline :api do
  plug(:accepts, ["json"])
  plug Api.Authentication # add this
end
# This is based on the Absinthe authentication documentation:
# https://hexdocs.pm/absinthe/context-and-authentication.html
defmodule Api.Authentication do
  @behaviour Plug

  import Plug.Conn

  def init(opts), do: opts

  def call(conn, _) do
    conn
    |> fetch_token()
    |> authorize_token()
    |> case do
      :ok -> conn
        |> put_private(:absinthe, %{context: %{authorized: true}})
      _ -> conn
        |> send_resp(401, "Unauthorized")
        |> halt
    end
  end

  def fetch_token(conn) do
    case get_req_header(conn, "authorization") do
      ["Bearer " <> token] -> token
      _ -> nil
    end
  end

  def authorize_token(nil), do: :error
  def authorize_token(""), do: :error

  def authorize_token(token) do
    case SecureCompare.compare(token, secret_token()) do
      true -> :ok
      _ -> :error
    end
  end

  def secret_token do
    System.get_env("YOUR_API_TOKEN")
  end
end

WebSocket Server

To secure the WebSocket connection, update user_socket:

defmodule MyAppWeb.UserSocket do
  ...

  def connect(%{"token" => token} = params, socket) do
    case SecureCompare.compare(token, Api.Authentication.secret_token()) do
      true ->
        {:ok, socket}
      _ ->
        :error
    end
  end
end

Client Query Validation

  • (using schema introspection result)

Query validation can be done at the client-side using schema introspection result to get closer to real integration tests without having to run a graphql server.

This can be done using the mix task:

$ mix graphql.validate_query -f schema.json <raw-query> $ mix graphql.validate_query -f schema.json $(cat <query.graphql-path>)

For more usage options try the help command:

$ mix graphql.validate_query -h

If you don't want to use the mix task, validation can be done at a module level by explicitly calling the static validator module:

schema_path = "path/to/schema.json"
query_string = "{ __schema { types { name } } }"
validation_strategy = :npm_graphql
CommonGraphqlClient.StaticValidator.validate(
  query_string,
  %{validation_strategy: validation_strategy,
    schema_path: schema_path}
)
# => :ok | {:error, error}

Schema validation can be done using validation strategies. The default validation strategy is using :npm-graphql. This requires npm and node binaries to be available (which is for most of the phoenix development environment)

For more information on this check out the documentation and examples for CommonGraphqlClient.StaticValidator.NpmGraphql

Using Npm

This uses npm and node commands to run schema validation. Make sure you have npm and node installed.

Using Native Elixir

This strategy will use native elixir for performing the validation. This is work in progress

Copyright and License

Copyright (c) 2018 Annkissam

This work is free. You can redistribute it and/or modify it under the terms of the MIT License. See the LICENSE.md file for more details.