A GraphQL server implementation in Clojure


Keywords
clojure, graphql
License
Other

Documentation

Lacinia

Clojars Project

CircleCI

Lacinia Manual | Lacinia Tutorial | API Documentation

This library is a full implementation of Facebook's GraphQL specification.

Lacinia should be viewed as roughly analogous to the official reference JavaScript implementation. In other words, it is a backend-agnostic GraphQL query execution engine. Lacinia is not an Object Relational Mapper ... it's simply the implementation of a contract sitting between the GraphQL client and your data.

Lacinia features:

  • An EDN-based schema language.

  • High performance parser for GraphQL queries, built on Antlr4.

  • Efficient and asynchronous query execution.

  • Full support for GraphQL types, interfaces, unions, enums, input objects, and custom scalars.

  • Full support for GraphQL subscriptions.

  • Full support of inline and named query fragments.

  • Full support for GraphQL Schema Introspection.

Lacinia has been developed with a set of core philosophies:

  • Prefer data over macros and other tricks. Compose your schema in whatever mix of data and code works for you.

  • Embrace Clojure: Use EDN data, keywords, functions, and persistent data structures.

  • Keep it simple: You provide the schema and a handful of functions to resolve data, and Lacinia does the rest.

  • Do the right thing: apply reasonable defaults without a lot of "magic".

This library can be plugged into any Clojure HTTP pipeline. The companion library lacinia-pedestal provides full HTTP support, including GraphQL subscriptions, for Pedestal.

Getting Started

For more detailed documentation, read the manual.

GraphQL starts with a schema definition of exposed types.

A schema starts as an EDN file; the example below demonstrates a small subset of the available options:

{:enums
 {:episode
  {:description "The episodes of the original Star Wars trilogy."
   :values [:NEWHOPE :EMPIRE :JEDI]}}

 :objects
 {:droid
  {:fields {:primary_functions {:type (list String)}
            :id {:type Int}
            :name {:type String}
            :appears_in {:type (list :episode)}}}

  :human
  {:fields {:id {:type Int}
            :name {:type String}
            :home_planet {:type String}
            :appears_in {:type (list :episode)}}}}

 :queries
 {:hero {:type (non-null :human)
         :args {:episode {:type :episode}}
         :resolve :get-hero}
  :droid {:type :droid
          :args {:id {:type String :default-value "2001"}}
          :resolve :get-droid}}}

A schema alone describes what data is available to clients, but doesn't identify where the data comes from; that's the job of a field resolver, provided by the :resolve key inside fields such as the :hero and :droid query.

The values here, :get-hero and :get-droid, are placeholders; the startup code of the application will use com.walmartlabs.lacinia.util/attach-resolvers to attach the actual field resolver function.

A field resolver is just a function which is passed the application context, a map of arguments values, and a resolved value from a parent field. The field resolver returns a value. If it's a scalar type, it should return a value that conforms to the defined type in the schema. If not, it's a type error.

The field resolver is totally responsible for obtaining the data from whatever external store you use: whether it is a database, a web service, or something else.

It's important to understand that every field has a field resolver, even if you don't define it explicitly. If you don't supply a field resolver, Lacinia provides a default field resolver, customized to the field.

Here's what the get-hero field resolver might look like:

(defn get-hero [context arguments value]
  (let [{:keys [episode]} arguments]
    (if (= episode :NEWHOPE)
      {:id 1000
       :name "Luke"
       :home_planet "Tatooine"
       :appears_in ["NEWHOPE" "EMPIRE" "JEDI"]}
      {:id 2000
       :name "Lando Calrissian"
       :home_planet "Socorro"
       :appears_in ["EMPIRE" "JEDI"]})))

In this greatly simplified example, the field resolver can simply return the resolved value. Field resolvers that return multiple values return a list, vector, or set of values.

In real applications, a field resolver might execute a query against a database, or send a request to another web service.

After attaching resolvers, it is necessary to compile the schema; this step performs validations, provide defaults, and organizes the schema for efficient execution of queries.

This needs only be done once, in application startup code:

(require '[clojure.edn :as edn]
         '[com.walmartlabs.lacinia.util :refer [attach-resolvers]]
         '[com.walmartlabs.lacinia.schema :as schema])

(def star-wars-schema
  (-> "schema.edn"
      slurp
      edn/read-string
      (attach-resolvers {:get-hero get-hero
                         :get-droid (constantly {})})
      schema/compile))

With the compiled application available, it can be used to execute requests; this typically occurs inside a Ring handler function:

(require '[com.walmartlabs.lacinia :refer [execute]]
         '[clojure.data.json :as json])

(defn handler [request]
  {:status 200
   :headers {"Content-Type" "application/json"}
   :body (let [query (get-in request [:query-params :query])
               result (execute star-wars-schema query nil nil)]
           (json/write-str result)})

Lacinia doesn't know about the web tier at all, it just knows about parsing and executing queries against a compiled schema. A companion library, lacinia-pedestal, is one way to expose your schema on the web.

User queries are provided as the body of a request with the content type application/graphql. The GraphQL query language is designed to look familiar to someone who is versant in JSON.

{
  hero {
    id
    name
  }
}

The execute function returns EDN data that can be easily converted to JSON. The :data key contains the value requested for the hero query in the request.

{:data
  {:hero {:id 2000
          :name "Lando Calrissian"}}}

This example request has no errors, and contained only a single query. GraphQL supports multiple queries in a single request. There may be errors executing the query, Lacinia will process as much as it can, and will report errors in the :errors key.

One of the benefits of GraphQL is that the client has the power to rename fields in the response:

{
  hero(episode: NEWHOPE) {
    movies: appears_in
  }
}
{:data {:hero {:movies [:NEWHOPE :EMPIRE :JEDI]}}}

This is just an overview, far more detail is available in the manual.

Status

Although this library is used in production at Walmart, it is still considered alpha software - subject to change. We expect to stabilize it in the near future.

To use this library with Clojure 1.8, you must include a dependency on clojure-future-spec.

More details are in the manual.

License

Copyright © 2017-2018 WalmartLabs

Distributed under the Apache License, Version 2.0.

Portions of the code are derived from the ordered and useful libraries, which are released under the terms of the Eclipse Public License - v 1.0.