Caliban

# Getting Started

Caliban is a purely functional library for creating GraphQL servers and clients in Scala.

For more details on Caliban Client, see the dedicated section. The rest of this page is about the backend part of the library.

The design principles of Caliban are the following:

# Dependencies

To use caliban, add the following line in your build.sbt file:

libraryDependencies += "com.github.ghostdogpr" %% "caliban" % "2.0.1"

The following modules are optional:

libraryDependencies += "com.github.ghostdogpr" %% "caliban-http4s"     % "2.0.1" // routes for http4s
libraryDependencies += "com.github.ghostdogpr" %% "caliban-akka-http"  % "2.0.1" // routes for akka-http
libraryDependencies += "com.github.ghostdogpr" %% "caliban-play"       % "2.0.1" // routes for play
libraryDependencies += "com.github.ghostdogpr" %% "caliban-zio-http"   % "2.0.1" // routes for zio-http
libraryDependencies += "com.github.ghostdogpr" %% "caliban-cats"       % "2.0.1" // interop with cats effect
libraryDependencies += "com.github.ghostdogpr" %% "caliban-monix"      % "2.0.1" // interop with monix
libraryDependencies += "com.github.ghostdogpr" %% "caliban-tapir"      % "2.0.1" // interop with tapir
libraryDependencies += "com.github.ghostdogpr" %% "caliban-federation" % "2.0.1" // interop with apollo federation

# A simple example

Creating a GraphQL API with Caliban is as simple as creating a case class. Indeed, the whole GraphQL schema will be derived from a case class structure (its fields and the other types it references), and the resolver is just an instance of that case class.

Let's say we have a class Character and 2 functions: getCharacters and getCharacter:

case class Character(name: String, age: Int)

def getCharacters: List[Character] = Nil
def getCharacter(name: String): Option[Character] = ???

Let's create a case class named Queries that will represent our API, with 2 fields named and modeled after the functions we want to expose (a record of functions). We then create a value of this class that calls our actual functions. This is our resolver.

// schema
case class CharacterName(name: String)
case class Queries(characters: List[Character],
                   character: CharacterName => Option[Character])
// resolver
val queries = Queries(getCharacters, args => getCharacter(args.name))

The next step is creating our GraphQL API definition. First, we wrap our query resolver inside a RootResolver, the root object that contains queries, mutations and subscriptions. Only queries are mandatory. Then we can call the graphQL function which will turn our simple resolver value into a GraphQL API definition. The whole schema will be derived at compile time, meaning that if it compiles, it will be able to serve it.

import caliban.GraphQL.graphQL
import caliban.RootResolver

val api = graphQL(RootResolver(queries))

You can use api.render to visualize the schema generated, in this case:

type Character {
  name: String!
  age: Int!
}

type Queries {
  characters: [Character!]!
  character(name: String!): Character
}

In order to process requests, you need to turn your API into an interpreter, which can be done easily by calling .interpreter. An interpreter is a light wrapper around the API definition that allows plugging in some middleware and possibly modifying the environment and error types (see Middleware for more info). Creating the interpreter may fail with a ValidationError if some type is found invalid.

for {
  interpreter <- api.interpreter
} yield interpreter

Now you can call interpreter.execute with a given GraphQL query, and you will get an ZIO[R, Nothing, GraphQLResponse[CalibanError]] as a response, with GraphQLResponse defined as follows:

case class GraphQLResponse[+E](data: ResponseValue, errors: List[E])

Use ResponseValue#toString to get the JSON representation of the result.

val query = """
  {
    characters {
      name
    }
  }"""

for {
  interpreter <- api.interpreter
  result      <- interpreter.execute(query)
  _           <- zio.ZIO.debug(result.data.toString)
} yield ()

A CalibanError can be:

  • a ParsingError: the query has invalid syntax
  • a ValidationError: the query was parsed but does not match the schema
  • an ExecutionError: an error happened while executing the query

Caliban itself is not tied to any web framework, you are free to expose this function using the protocol and library of your choice. The caliban-http4s (opens new window) module provides an Http4sAdapter that exposes an interpreter over HTTP and WebSocket using http4s. There are also similar adapters for Akka HTTP, Play and zio-http. Read more on the adapters' documentation.

Combining GraphQL APIs

You don't have to define all your root fields into a single case class: you can use smaller case classes and combine GraphQL objects using the |+| operator.

val api1 = graphQL(...)
val api2 = graphQL(...)

val api = api1 |+| api2

You can use .rename to change the names of the generated root types.

# Mutations

Creating mutations is the same as queries, except you pass them as the second argument to RootResolver:

import zio.Task

case class CharacterArgs(name: String)
case class Mutations(deleteCharacter: CharacterArgs => Task[Boolean])
val mutations = Mutations(_ => ???)
val api = graphQL(RootResolver(queries, mutations))

# Subscriptions

Similarly, subscriptions are passed as the third argument to RootResolver:

import zio.stream.ZStream

case class Subscriptions(deletedCharacter: ZStream[Any, Nothing, Character])
val subscriptions = Subscriptions(???)
val api = graphQL(RootResolver(queries, mutations, subscriptions))

All the fields of the subscription root case class MUST return ZStream or ? => ZStream objects. When a subscription request is received, an output stream of ResponseValue (a StreamValue) will be returned wrapped inside an ObjectValue.

Schemas →