eikek / sbt-openapi-schema

Generate schema sources for Scala, Java and Elm from an openapi 3.0 spec.

Version Matrix

SBT OpenApi Schema Codegen

Build Status Scala Steward badge License

This is an sbt plugin to generate Scala or Elm code given an openapi 3.x specification. Unlike other codegen tools, this focuses only on the #/components/schema section. Also, it generates immutable classes and optionally the corresponding JSON conversions.

  • Scala: case classes are generated and JSON conversion via circe.
  • Elm: records are generated and constructors for "empty" values. It works only for objects. JSON conversion is generated using Elm's default encoding support and the json-decode-pipeline module for decoding.
  • JSON support is optional.

The implementation is based on the swagger-parser project.

It is possible to customize the code generation.

Usage

Add this plugin to your build in project/plugins.sbt:

addSbtPlugin("com.github.eikek" % "sbt-openapi-schema" % "0.5.0")

Please check the git tags or maven central for the current version. Then enable the plugin in some project:

enablePlugins(OpenApiSchema)

There are two required settings: openapiSpec and openapiTargetLanguage. The first defines the openapi.yml file and the other is a constant from the Language object:

import com.github.eikek.sbt.openapi._

project.
  enablePlugins(OpenApiSchema).
  settings(
    openapiTargetLanguage := Language.Scala
    openapiSpec := (Compile/resourceDirectory).value/"test.yml"
  )

The sources are automatically generated when you run compile. The task openapiCodegen can be used to run the generation independent from the compile task.

Configuration

The configuration is specific to the target language. There exists a separate configuration object for Scala and Elm.

The key openapiScalaConfig defines some configuration to customize the code generation.

For Scala, it looks like this:

case class ScalaConfig(
    mapping: CustomMapping = CustomMapping.none,
    json: ScalaJson = ScalaJson.none
) {

  def withJson(json: ScalaJson): ScalaConfig =
    copy(json = json)

  def addMapping(cm: CustomMapping): ScalaConfig =
    copy(mapping = mapping.andThen(cm))

  def setMapping(cm: CustomMapping): ScalaConfig =
    copy(mapping = cm)
}

By default, no JSON support is added to the generated classes. This can be changed via:

openapiScalaConfig := ScalaConfig().withJson(ScalaJson.circeSemiauto)

This generates the encoder and decoder using circe. Note, that this plugin doesn't change your libraryDependencies setting. So you need to add the circe dependencies yourself.

The CustomMapping class allows to change the class names or use different types (for example, you might want to change LocalDate to Date).

It looks like this:

trait CustomMapping { self =>

  def changeType(td: TypeDef): TypeDef

  def changeSource(src: SourceFile): SourceFile

  def andThen(cm: CustomMapping): CustomMapping = new CustomMapping {
    def changeType(td: TypeDef): TypeDef =
      cm.changeType(self.changeType(td))

    def changeSource(src: SourceFile): SourceFile =
      cm.changeSource(self.changeSource(src))
  }
}

There are convenient constructors in its companion object.

It allows to use different types via changeType or change the source file. Here is a build.sbt example snippet:

import com.github.eikek.sbt.openapi._

val CirceVersion            = "0.14.1"
libraryDependencies ++= Seq(
  "io.circe" %% "circe-generic" % CirceVersion,
  "io.circe" %% "circe-parser"  % CirceVersion
)

openapiSpec := (Compile/resourceDirectory).value/"test.yml"
openapiTargetLanguage := Language.Scala
Compile/openapiScalaConfig := ScalaConfig()
    .withJson(ScalaJson.circeSemiauto)
    .addMapping(CustomMapping.forType({ case TypeDef("LocalDateTime", _) =>
      TypeDef("Timestamp", Imports("com.mypackage.Timestamp"))
    }))
    .addMapping(CustomMapping.forName({ case s => s + "Dto" }))

enablePlugins(OpenApiSchema)

It adds circe JSON support and changes the name of all classes by appending the suffix "Dto". It also changes the type used for local dates to be com.mypackage.Timestamp.

Elm

There is some experimental support for generating Elm data structures and corresponding JSON conversion functions. When using the decodePipeline json variant, you need to install these packages:

elm install elm/json
elm install NoRedInk/elm-json-decode-pipeline

The default output path for elm sources is target/elm-src. So in your elm.json file, add this directory to the source-directories list along with the main source dir. It may look something like this:

{
    "type": "application",
    "source-directories": [
        "modules/webapp/target/elm-src",
        "modules/webapp/src/main/elm"
    ],
    "elm-version": "0.19.0",
    "dependencies": {
        "direct": {
            "NoRedInk/elm-json-decode-pipeline": "1.0.0",
            "elm/browser": "1.0.1",
            "elm/core": "1.0.2",
            "elm/html": "1.0.0",
            "elm/json": "1.1.3"
        },
        "indirect": {
            "elm/time": "1.0.0",
            "elm/url": "1.0.0",
            "elm/virtual-dom": "1.0.2"
        }
    },
    "test-dependencies": {
        "direct": {},
        "indirect": {}
    }
}

It always generates type aliases for records.

While source files for scala are added to sbt's sourceGenerators so that they get compiled with your sources, the elm source files are not added anywhere, because there is no support for Elm in sbt. However, in the build.sbt file, you can tell sbt to generate the files before compiling your elm app. This can be configured to run during resource generation. Example:

// Put resulting js file into the webjar location
def compileElm(logger: Logger, wd: File, outBase: File, artifact: String, version: String): Seq[File] = {
  logger.info("Compile elm files ...")
  val target = outBase/"META-INF"/"resources"/"webjars"/artifact/version/"my-app.js"
  val proc = Process(Seq("elm", "make", "--output", target.toString) ++ Seq(wd/"src"/"main"/"elm"/"Main.elm").map(_.toString), Some(wd))
  val out = proc.!!
  logger.info(out)
  Seq(target)
}

val webapp = project.in(file("webapp")).
  enablePlugins(OpenApiSchema).
  settings(
    openapiTargetLanguage := Language.Elm,
    openapiPackage := Pkg("Api.Model"),
    openapiSpec := (Compile/resourceDirectory).value/"openapi.yml",
    openapiElmConfig := ElmConfig().withJson(ElmJson.decodePipeline),
    Compile/resourceGenerators += (Def.task {
      openapiCodegen.value // generate api model files
      compileElm(streams.value.log
        , (Compile/baseDirectory).value
        , (Compile/resourceManaged).value
        , name.value
        , version.value)
    }).taskValue,
    watchSources += Watched.WatchSource(
      (Compile/sourceDirectory).value/"elm"
        , FileFilter.globFilter("*.elm")
        , HiddenFileFilter
    )
  )

This example assumes a elm.json project file in the source root.

Discriminator Support

OpenAPI 3.0 enables to introduce subtyping on generated schemas by using discriminators.

Two of these are currently supported only in Scala : oneOf and allOf.

Setup

In order to provide JSON conversion for these discriminators with Circe, we need to make use of circe-generic-extras

An example build.sbt using the plugin would look like the following:

import com.github.eikek.sbt.openapi._

libraryDependencies ++= Seq(
  "io.circe" %% "circe-generic-extras" % "0.11.1",
  "io.circe" %% "circe-core" % "0.11.1",
  "io.circe" %% "circe-generic" % "0.11.1",
  "io.circe" %% "circe-parser" % "0.11.1"
)

openapiSpec := (Compile/resourceDirectory).value/"test.yml"
openapiTargetLanguage := Language.Scala
openapiScalaConfig := ScalaConfig().
  withJson(ScalaJson.circeSemiautoExtra).
  addMapping(CustomMapping.forName({ case s => s + "Dto" }))

enablePlugins(OpenApiSchema)

Handle allOf keywords in Scala

Here is an example OpenAPI spec and the resulting Scala models with JSON conversions

components:
  schemas:
    Pet:
      type: object
      discriminator:
        propertyName: petType
      properties:
        name:
          type: string
        petType:
          type: string
      required:
      - name
      - petType
    Cat:  ## "Cat" will be used as the discriminator value
      description: A representation of a cat
      allOf:
      - $ref: '#/components/schemas/Pet'
      - type: object
        properties:
          huntingSkill:
            type: string
            description: The measured skill for hunting
        required:
        - huntingSkill
    Dog:  ## "Dog" will be used as the discriminator value
      description: A representation of a dog
      allOf:
      - $ref: '#/components/schemas/Pet'
      - type: object
        properties:
          packSize:
            type: integer
            format: int32
            description: the size of the pack the dog is from
        required:
        - packSize
import io.circe._
import io.circe.generic.extras.semiauto._
import io.circe.generic.extras.Configuration

sealed trait PetDto {
  val name: String
}
object PetDto {
  implicit val customConfig: Configuration = Configuration.default.withDefaults.withDiscriminator("petType")

  case class Cat (
    huntingSkill: String, name: String
  ) extends PetDto

  case class Dog (
    packSize: Int, name: String
  ) extends PetDto

  object Cat {
    implicit val customConfig: Configuration = Configuration.default.withDefaults.withDiscriminator("petType")
    private implicit val jsonDecoder: Decoder[Cat] = deriveDecoder[Cat]
    private implicit val jsonEncoder: Encoder[Cat] = deriveEncoder[Cat]
  }

  object Dog {
    implicit val customConfig: Configuration = Configuration.default.withDefaults.withDiscriminator("petType")
    private implicit val jsonDecoder: Decoder[Dog] = deriveDecoder[Dog]
    private implicit val jsonEncoder: Encoder[Dog] = deriveEncoder[Dog]
  }

  implicit val jsonDecoder: Decoder[PetDto] = deriveDecoder[PetDto]
  implicit val jsonEncoder: Encoder[PetDto] = deriveEncoder[PetDto]
}

Notes about the above example:

  • The internal schemas ("Dog" and "Cat") have private encoder/decoders so that they are only encoded and decoded as the trait interface. If you try to decode as a Dog or Cat type, the circe-generic-extras doesn't include the discriminant type
  • The mapping functionality (adding "Dto") is only used on the sealed trait since the discriminant type uses the name of the inner case classes ("Dog" and "Cat").

Handle oneOf keywords in Scala

Another way of transform composed schemas into sealed trait hierarchies is to use oneOf.

Pet:
  type: object
  discriminator:
    propertyName: petType
  oneOf:
    - $ref: '#/components/schemas/Cat'
    - $ref: '#/components/schemas/Dog'
Cat:  ## "Cat" will be used as the discriminator value
  description: A representation of a cat
  properties:
    huntingSkill:
      type: string
      description: The measured skill for hunting
      enum:
        - clueless
        - lazy
        - adventurous
        - aggressive
    name:
      type: string
    petType:
      type: string
  required:
    - huntingSkill
    - name
    - petType
Dog:  ## "Dog" will be used as the discriminator value
  description: A representation of a dog
  properties:
    packSize:
      type: integer
      format: int32
      description: the size of the pack the dog is from
      default: 0
      minimum: 0
    name:
      type: string
    petType:
      type: string
  required:
    - packSize
    - name
    - petType
import io.circe._
import io.circe.generic.extras.semiauto._
import io.circe.generic.extras.Configuration

sealed trait PetDto {
} 
object PetDto {
  implicit val customConfig: Configuration = Configuration.default.withDefaults.withDiscriminator("petType")
  case class Cat (
    huntingSkill: String, name: String, petType: String
  ) extends PetDto
  case class Dog (
    packSize: Int, name: String, petType: String
  ) extends PetDto
  object Cat {
    implicit val customConfig: Configuration = Configuration.default.withDefaults.withDiscriminator("petType")
    private implicit val jsonDecoder: Decoder[Cat] = deriveDecoder[Cat]
    private implicit val jsonEncoder: Encoder[Cat] = deriveEncoder[Cat]
  }
  object Dog {
    implicit val customConfig: Configuration = Configuration.default.withDefaults.withDiscriminator("petType")
    private implicit val jsonDecoder: Decoder[Dog] = deriveDecoder[Dog]
    private implicit val jsonEncoder: Encoder[Dog] = deriveEncoder[Dog]
  }
  implicit val jsonDecoder: Decoder[PetDto] = deriveDecoder[PetDto]
  implicit val jsonEncoder: Encoder[PetDto] = deriveEncoder[PetDto]
} 

Unlike allOf, oneOf doesn't permit subschemas to inherit fields from their parent. This kind of relation fits well to algebraic data types encodings in Scala.

Static Documentation

The plugin can run the swagger codegen tool or redoc to produce a static HTML page of the OpenAPI specification file.

Define which generator to use via:

openapiStaticGen := OpenApiDocGenerator.Redoc //or
openapiStaticGen := OpenApiDocGenerator.Swagger

Note that nodejs (the npx command) is required for redoc! The default is swagger.

Then use the openapiStaticDoc task to generate the documentation from your openapi specification.

Additionally, there is also a task that runs openapi-cli lint against your specification file. This also requires to have nodejs installed.

Credits

First, thank you all who reported issues! It follows a list of contributions in form of code. If you find yourself missing, please let me know or open a PR.

  • @xela85 for adding oneOf keywords support (#42)
  • @mhertogs for adding support for discriminators (#8)