Skip to content
This repository has been archived by the owner on Apr 10, 2019. It is now read-only.

zalando-stups/play-swagger

Repository files navigation

The Play-Swagger plugin is now renamed api-first-hand. This version is no longer under active development.

Api-First-Hand is actively mantained and offers full functionality of Play-Swagger with an exception of Play 2.4 support. Please navigate to api-first-hand if you'd like to check out Play-Swagger or create an issue.

--

--

Play-Swagger

Build Status codecov Gitter Chat

Compatibility

  • Play 2.4
  • Swagger (OpenAPI) 2.0

Status

This plugin should be enabled using the play-swagger-service activator template as the version in this repository is under active development. The status of this software is beta, an end-to-end functional release intended to demonstrate the possibility to generate following from a Swagger specification:

  • Play route files
  • Generators of random test data
  • Wrappers for Play route files to convert semantics from http-related to domain-related (controller_base)
  • Skeletons for the domain-driven controller implementation
  • Model classes and validation rules
  • Unit tests for invalid and valid parameter sets
  • Security extractors (if needed)
  • Skeletons for custom deserializers (if needed)

We benefit from community feedback. All comments are welcome!

Play-Swagger Tutorial

This tutorial is based on the play-swagger-service activator template.

$ activator new playground play-swagger-service

The template project contains following:

  • tutorial folder with HTML tutorial
  • public/swagger folder containing static files needed for swagger UI
  • project folder containing pre-configured plugins.sbt file with a definition of all required resolvers and plugins
  • conf folder with following customized contents:
    • routes file with route configuration for Swagger UI, example specification and commented out links to other examples
    • example.yaml, a demo Swagger specification. The specification has a dummy implementation in app folder.
    • examples folder containing other different Swagger specification examples. Each specification in this folder represents some aspect of the Play-Swagger plugin in more details. For the specification to be picked up by the plugin it must be moved into the conf folder. It is allowed to have multiple Swagger specifications in the conf folder at the same time.
  • app directory with following template implementations:
    • controllers/Swagger.scala - a backend side of the Swagger UI
    • generated_controllers/example.yaml.scala - a dummy implementation of the example controller. Will be (re)generated if deleted
    • security/example.yaml.scala - a marshaller for OAuth2 tokens. Will not be regenerated until a) deleted or renamed b) explicitly requested by issuing a apiFirstSecurity command

Welcome to Play-Swagger

Congratulations, you just created a new Play-Swagger application!

The Play Framework with the Play-Swagger plugin make it easy to build RESTful web services from a Swagger API specification as the single source of truth. Play is based on a lightweight, stateless, web-friendly architecture. Built on Akka, Play provides predictable and minimal resource consumption for highly-scalable applications. The Play-Swagger plugin takes Swagger API definitions and treats them as the single source of truth of your REST services.

Play-Swagger supports round-trip regeneration and compilation of:

  • Play routes definitions (managed).
  • Swagger domain model definitions and parameters onto Scala case classes (managed).
  • Swagger domain model constraints onto Play validations (managed).
  • Generators for random test data generation of parameter values (managed).
  • Unit tests for validating your service at the API boundary (managed).
  • Swagger path definitions onto skeletons for Play controller implementations (unmanaged).

In the list above, "(managed)" means that the code is managed by sbt. The code is not controlled and altered by you, the programmer of the REST service. The plugin takes your Swagger API definition as the single source of truth and regenerates these code parts in a consistent manner. You'll instead be focusing on implementing the service business logic in an (unmanaged) Play controller class that is generated once. Subsequent regenerations keep the code that you have added, either by commenting out the parts that are no longer valid, or by adding parts that are needed because you have made a change to the API.

Manual generation and compilation of:

  • Security extractors
  • Unmarshallers for custom content types

is supported in the case if

a) No security extractor or unmarshaller with the same name already exists b) The developer issues apiFirstSecurity or apiFirstMarshallers sbt command

Run Your Application

Before we go any further, let's run the application.

  • Open a shell and cd into your service project directory.
  • Start sbt and run the service.
  • View the running application at http://localhost:9000.

The service template comes with the Swagger UI frontend included, run statically from the within Play, which provides a sandbox for your service. The template is configured with a template Swagger API definition called example.yaml and located in the conf directory of the Play application.

The example.yaml definition provides an example API description

This definition contains three end points:

  • the /token path, which accept the GET and POST methods
  • the /todos/{user_id}, which accepts the GET method.

The GET /token API plays a role of an authentication server and is used by the Swagger UI for OAuth token requests. The POST /token API represents an authorization server and is used by the security part of the generated code to validate OAuth tokens.

The GET /todos/{user_id} takes a path parameter user_id and returns a TODO list for given user. For the client to be allowed to access this endpoint, it must provide an OAuth token with the scope admin:org. The token can be requested using the Swagger UI.

Try it out for yourself:

Click the default button to expand the API definition in the Swagger UI.

Play Routes Integration

As a Play application developer, you are used to defining your endpoints in the conf/routes file. Not so with the Play-Swagger plugin! Swagger API specifications already define endpoints as path definitions, as seen in the example above. So why do the work twice, right? Instead, the Play-Swagger plugin requires you to link your API definition in the routes file ones—making all Swagger API-defined endpoints available as children of one single path context location, and generating Play route definitions from them (as shown below):

->      /example        example.yaml.Routes

Note that the conf/routes file provided by this activator template also contains a couple of additional GET mappings required for the the Swagger UI sandbox.

There are a couple of commented out links to other examples. If you activate some specification by moving it from the examples folder into the conf folder, you'll need to uncomment an appropriate line in the routes file in order for play to be able to find it.

Swagger Domain Definitions

Scala domain model definitions are generated for all data types defined as Swagger parameters in an API specification. Swagger parameters can be of path, query, header, form or body types, and consist of either primitive data types or more complex types composed from objects and arrays with primitives as leaves.

Both primitive types and complex types are mapped to scala.

As an example, let's look at the Swagger API specification file simple.petstore.api.yaml, which defines the API of a simple pet store. It contains a model definition for a pet.

definitions:
  pet:
    required:
      - id
      - name
    properties:
      id:
        type: integer
        format: int64
      name:
        type: string
      tag:
        type: string

This definition consists of an object pet containing the required properties id and name and the optional property tag. The Swagger primitive types of these properties are a 64-bit integer and (twice) a string, successively. The Play-Swagger plugin will map this definition on to a generated Scala model.

package simple.petstore.api

package object yaml {

    type PetTag = Option[String]

    case class Pet(id: Long, name: String, tag: PetTag)
}

This generated model contains a type definition PetTag, which declares a type alias for the optional tag property, and a Pet case class with the properties as named in the Swagger API definition and mapped on the subsequent Scala primitive or declared types. The case class and type alias are generated in an package object yaml, this package object itself is contained in the package simple.petstore.api so that full object name corresponds to the API filename.

Note that models are generated within a Play application as managed code in the target folder. Generated model code is not intended to be altered. We should instead look upon the Swagger definition as the single source of truth, and as the source code that defines our model. The Swagger specification file of our API is, in that sense, part of the codebase. Even though the generated Pet case class is managed by the plugin, and not us, it can (of course) be used in our application codebase after being imported.

import simple.petstore.api.yaml._

val pet = Pet(0L, "Tucker", Some("Greyhound"))

Specification Cross-References

A $ref element of the specification is allowed to contain a name of file as it's part. Because of this, it is possible to split a single specification into multiple files as shown in cross_spec_references.yaml example. It is also possible to reference a definition in one specification from another specification. In this case for each reference an independent copy of the class definition will be created for each referencing specification. The definition is then placed into the appropriate package for each specification.

Thus, even if multiple classes with the same name and structure might be generated, they all will coexist in their own separate namespaces and won't be interchangeable.

Primitive Types

Swagger version 2.0 allows for primitive data types based on the types defined by JSON-Schema.

When generated as Scala, the following mapping applies:

Common Name Swagger Type Swagger Format Scala Type
integer integer int32 scala.Int
long integer int64 scala.Long
float number float scala.Float
double number double scala.Double
big int integer scala.math.BigInt
big decimal number scala.math.BigDecimal
boolean boolean scala.Boolean
string string scala.String
byte string byte de.zalando.play.controllers.Base64String
binary string binary de.zalando.play.controllers.BinaryString
date string date org.joda.time.LocalDate
datetime string date-time org.joda.time.DateTime
password string password scala.String
file file java.io.File

Additionally, if a validation of type "enum" is defined for some primitive type, a trait and a set of case objects forming an ADT will be generated for this enum.

Complex Types

Complex types are made up of primitive objects, or nested objects.

Objects

Complex object types are defined in Swagger model definitions as either objects or arrays.

Objects are, again, based on the JSON-Schema specification and defined as Swagger Schema Objects for parameter definitions of type: "object". For example, given a Swagger API definition file api.yaml containing a model that defines a person as an object with the properties name and age of the primitive types string and integer subsequently, this object will be mapped on a Scala case class, and generated in a Scala package object (namespace) with the same name as the extension of the file the specification is read from and in a package with the same name as the Swagger definition file in which the model is defined—that is, api

definitions:
  person:
    type: object
    required:
      - name
      - age
    properties:
      name:
        type: string
      age:
        type: integer
        format: int32

Is generated into:

package api
package object yaml {
    case class Person(name: String, age: Int) 
}

Nested Objects

Nested objects are generated adjourned but referenced hierarchically. E.g.

definitions:
  parent:
    type: object
    required:
      - child
    properties:
      child:
        type: object
        required:
          - name
        properties:
          name:
            type: string

Is generated into:

package api
package object yaml {
    case class Parent(child: ParentChild) 
    case class ParentChild(name: String) 
}

Optionality

Swagger, by default, defines object properties to be optional, which can be overridden by providing a list of required object properties as already used in the examples above. Optional properties are mapped upon Scala's Option type, for which a type alias is generated for each property that is optional. E.g.

definitions:
  product:
    required:
      - name
    properties:
      name:
        type: string
      tag:
        type: string

Which is generated as:

package api
package object yaml {
    type ProductTag = Option[String]
    case class Product(name: String, tag: ProductTag) 
}

As objects can be nested, so can object property optionality. To facilitate for nested optionality, we generate a nested scala Option type alias. E.g.

definitions:
  Basic:
    properties:
      optional:
        type: object
        properties:
          nested:
            type: string

Which is generated as:

package api
package object yaml {
    type BasicOptional = Option[BasicOptionalOpt]
    type BasicOptionalNested = Option[String]

    case class BasicOptionalOpt(nested: BasicOptionalNested) 
    case class Basic(optional: BasicOptional) 
}

Parameter optionality

As object properties can be optional, so can be query, header, body or form parameters. In the case if they are not required, they are mapped to the Scala's Option type.

Path parameters are must be declared as required.

In the case, if a parameter is not required, it is allowed to have a default value.

Extension

Objects can extend other objects via employment of Swagger's allOff property. In the example below, the ExtendedErrorModel inherits all of the properties of the ErrorModel which it refers to—that is, the properties message and code—and extends this model with the property rootCause. Swagger object extension is mapped by duplicating inherited properties in the object that extends. E.g.

definitions:
  ErrorModel:
    type: object
    required:
    - message
    - code
    properties:
      message:
        type: string
      code:
        type: integer
  ExtendedErrorModel:
    allOf:
    - $ref: '#/definitions/ErrorModel'
    - type: object
      required:
      - rootCause
      properties:
        rootCause:
          type: string

Which is generated as:

package api
package object yaml {
  import scala.math.BigInt
  case class ErrorModel(message: String, code: BigInt) 
  case class ExtendedErrorModel(message: String, code: BigInt, rootCause: String) 
}

Polymorphism

Polymorphic object definitions are possible through employment of the Swagger discriminator property. In the example definition below, an abstract Pet defines what concrete Cat and Dogs have in common. Swagger object models define data, so a discriminator property is required to distinguish concrete cat and dog instances as they are serialised to and from the API. In this sense, the discriminator property works in the same way as a discriminator column works in ORM frameworks when mapping a class hierarchy onto a single table. It simply contains a value that maps onto one of the concrete types—for example, petType: "Cat" or petType: "Dog".

definitions:
  Pet:
    discriminator: petType
    properties:
      name:
        type: string
      petType:
        type: string
    required:
    - name
    - petType
  Cat:
    allOf:
    - $ref: '#/definitions/Pet'
    - properties:
        huntingSkill:
          type: string
          default: lazy
          enum:
          - clueless
          - lazy
          - adventurous
          - aggressive
      required:
      - huntingSkill
  Dog:
    allOf:
    - $ref: '#/definitions/Pet'
    - properties:
        packSize:
          type: integer
          format: int32
      required:
      - packSize

Which is generated as:

package api

package object yaml {

    trait IPet {
        def name: String
        def petType: String
    }

    case class Cat(name: String, petType: String, huntingSkill: CatHuntingSkill) extends IPet
    case class Dog(name: String, petType: String, packSize: Int) extends IPet
    case class Pet(name: String, petType: String) extends IPet

    sealed trait CatHuntingSkill { def value: String }
    case object Clueless extends CatHuntingSkill { val value = "clueless" }
    case object Lazy extends CatHuntingSkill { val value = "lazy" }
    case object Adventurous extends CatHuntingSkill { val value = "adventurous" }
    case object Aggressive extends CatHuntingSkill { val value = "aggressive" }
    implicit def stringToCatHuntingSkill(in: String): CatHuntingSkill = in match {
        case "clueless" => Clueless
        case "lazy" => Lazy
        case "adventurous" => Adventurous
        case "aggressive" => Aggressive
    }
}

Please note how the enumeration of cat's huntingSkill's get's translated into the ADT with a sealed trait CatHuntingSkill and four case objects implementing that trait.

Additional Properties

Swagger's model language allows objects' additional properties to be loosely defined employing the additionalProperties annotation in order to model dictionaries. These dictionaries are mapped to Scala's Map type, for which a type alias is generated following the same (by now) well-known pattern as for optional properties, with the map's key parameter type being a Scala String.

A Swagger additional property definition takes as its type property the element type of the dictionary, which can be of primitive or complex type and which is mapped on Scala as the map's value parameter type. Swagger allows for one additionalProperties annotation per object definition, so we can generate this Scala parameter with the static name additionalProperties.

In the following example we define a Swagger model object definition KeyedArray that uses the additionalProperties annotation to provide the object with a set of key value mappings from string to array. E.g.

definitions:
  KeyedArrays:
    type: object
    additionalProperties:
      type: array
      items:
        type: integer

Which is generated as:

package api

package object yaml {

    import de.zalando.play.controllers.ArrayWrapper
    import scala.math.BigInt
    import scala.collection.immutable.Map

    type KeyedArraysAdditionalPropertiesCatchAll = ArrayWrapper[BigInt]
    type KeyedArraysAdditionalProperties = Map[String, KeyedArraysAdditionalPropertiesCatchAll]
    case class KeyedArrays(additionalProperties: KeyedArraysAdditionalProperties) 
}

Arrays

Swagger's array is used to define properties that hold sets or lists of model values—possibly of a primitive type, but complex element types are also allowed. Depending on the place where the array definition appears, Swagger array can be mapped to one of two Scala types, parametrised for the element type that it contains:

  • if an array only defined inline as a part of the response definition, it is translated to a Seq type
  • otherwise (array appears in the parameter definition or in the definitions part of the specification) it is defined as a de.zalando.play.controllers.ArrayWrapper

For example, in the snippet below, an Activity object definition is referred to as an item element in the messages property of type: array of the containing object definition Example. A Scala type alias will be generated for the array type (just as we've seen before with optional properties), after which the array-containing property can be generated within the case class as being of this alias type. E.g. in the Swagger definition and code

definitions:
  Activity:
    type: object
    required:
    - actions
    properties:
      actions:
        type: string
  Example:
    type: object
    required:
    - messages
    properties:
      messages:
        type: array
        items:
          $ref: '#/definitions/Activity'

Which is generated as:

package api

package object yaml {

    import de.zalando.play.controllers.ArrayWrapper

    type ExampleMessages = ArrayWrapper[Activity]

    case class Activity(actions: String) 
    case class Example(messages: ExampleMessages) 
}

If the description of the same array is inlined as a part of the response definition like that:

paths:
  /api:
    get:
      responses:
        200:
          schema:
            type: object
            required:
            - messages
            properties:
              messages:
                type: array
                items:
                  $ref: '#/definitions/Activity'
          description: array payload
definitions:
  Activity:
    type: object
    required:
    - actions
    properties:
      actions:
        type: string

than the Seq scala type will be used:

package api
package object yaml {
    type ApiGetResponses200Messages = Seq[Activity]
    case class Activity(actions: String) 
    case class ApiGetResponses200(messages: ApiGetResponses200Messages) 
}

Nested Arrays

Array definition types can be nested and are possibly optional. The following (contrived) snippet depicts the generated Scala code when both definition types are employed in a somewhat non-useful manner. The intent of this example is to show that the case class definitions are rather concisely generated, even though a stack of type aliases is needed to make sure that we still refer in Scala code to an aptly named Swagger definition—especially in conjunction with the object properties being optional. Next to its benefits, type safety against null pointers does have an associated cost as well.

definitions:
  Activity:
    type: object
    properties:
      actions:
        type: string
  Example:
    type: object
    properties:
      messages:
        type: array
        items:
          type: array
          items:
            $ref: '#/definitions/Activity'
      nested:
        type: array
        items:
          type: array
          items:
            type: array
            items:
              type: array
              items:
                type: string

Which is generated as:

package api

package object yaml {

    import de.zalando.play.controllers.ArrayWrapper

    type ExampleMessagesOpt = ArrayWrapper[ExampleMessagesOptArr]
    type ExampleMessages = Option[ExampleMessagesOpt]
    type ExampleNested = Option[ExampleNestedOpt]
    type ExampleMessagesOptArr = ArrayWrapper[Activity]
    type ExampleNestedOptArrArrArr = ArrayWrapper[String]
    type ExampleNestedOptArrArr = ArrayWrapper[ExampleNestedOptArrArrArr]
    type ActivityActions = Option[String]
    type ExampleNestedOptArr = ArrayWrapper[ExampleNestedOptArrArr]
    type ExampleNestedOpt = ArrayWrapper[ExampleNestedOptArr]

    case class Activity(actions: ActivityActions) 
    case class Example(messages: ExampleMessages, nested: ExampleNested) 
}

Swagger Validations

Swagger API definitions allow for constraints to be put on parameter types. We have already seen the required constraint, used to mark a parameter or specific field within a domain definition to be required upon input. Additional constraints, as defined by the Parameter Object, can be added to your API definition. The Play-Swagger plugin will generate validations for these parameter constraints and make sure that your controller methods are only called if the input of your service complies to those constraints.

In the example below, the API definition of the token parameter of type Base64String, as the form parameter, contains validation rules for the lenght of the perameter as well as a regexp pattern the value of the parameter must confirm to. The parameter is also required.

...
parameters:
      - name: token
        in: formData
        description: oauth2 token
        type: string
        format: byte
        pattern: "[A-Za-z0-9]*"
        minLength: 5
        maxLength: 100
        required: true
...

Let's take another example:

...
    get:
      parameters:
      - name: state
        in: query
        description: Any application state to be forwarded back to the frontend
        type: string
        minLength: 1
        maxLength: 110
        required: false
...

The state parameter is of type string, is not required and has no default value. It is also only allowed to have a state of length between 1 and 110, otherwise it won't pass validation. For the demo purposes, let's change it's type to integer and make it required.

As the parameter is required now, the default value cannot be present. The maxLength and maxLength validations are not allowed for integer parameters, therefore let's replace them with minimum and maximum values:

...
    get:
      parameters:
      - name: state
        in: query
        description: Any application state to be forwarded back to the frontend
        type: integer
        format: int32
        required: true
        minimum: 2000
        maximum: 2100      
...

As we just changed the parameter type, refreshing Swagger UI will, in addition to generating validations for that parameter type, also force a regeneration of the model consistent with the validation. That's nice, but note that it will break the current implementation of the controller class, as the implementation of the postAction expects state to be of type String.

Validation screenshot

Let's change the implementation. The second parameter state is no longer of type Option[String] but of type Int. We change the implementation to take this fact into the account:

...
val tokenGet = tokenGetAction { input: (String, String, String, Int) =>
    val (redirect_uri, scope, response_type, state) = input
    // ----- Start of unmanaged code area for action  TokenService.tokenGet
    val statePart = s"""state=$state"""
...
}

Refreshing Swagger UI and trying out a couple of integer values for state shows that the service now excepts value within the range [2000..2100], but returns a descriptive error when outside. I.e.

[
  {
    "messages": [
      "error.max"
    ],
    "args": [
      2100
    ]
  }
]

Test Generators

Having an API definition as the single source of truth in your codebase—with formal type specification of the in- and output values, including their constraints—provides for a powerful feature when it comes to testing. The Play-Swagger plugin automates the creation of test data generators that can drive property checks directly from the API specification. Play-Swagger derives data generators and unit tests directly from your Swagger API specification.

Property-based testing using generator-driven property checks is a cool way to test the validity of your application according to the rules or properties that apply to your application. Properties, in this sense, are high-level specifications that should always hold for a range of data values. The idea is to generate a range of data values for your data types and let (also generated) tests assert that the properties of these data types hold. A Swagger API definition contains formal type definitions and constraints for all data values, and the Play-Swagger plugin maps these types on managed Scala source code that represents the data types, so it is also possible to map these API definitions on test data generators that provide a range of data values for these types. The plugin does exactly that: It creates managed test data generators and unit tests that assert whether your application still complies to your specification. It does so in a single-source-of-truth manner, taking the Swagger API definition as the source.

We employ the ScalaTest property-based testing functionality as the framework to generate the data values, and map the data types of our API definition on the test data generators that are created by the plugin. ScalaTest provides org.scalacheck.Gen and org.scalacheck.Arbitrary objects with utility methods that help generate a range of (possibly arbitrary) data values for common Scala types and primitives. The Play-Swagger plugin uses these methods to create test data generators specific for the data types of our API definition. When necessary, it composes generators from primitive types into generators for complex types, so that you end up with a set of generators that provide test data for your complete API.

As an example, let's take the API definition for the simple pet store—trimmed down to the parts defining parameter types, and (for brevity) omitting any non-data definitions and error definitions:

paths:
  /pets:
    get:
      parameters:
        - name: limit
          in: query
          required: false
          type: integer
          format: int32
      responses:
        default:
          description: error payload
    post:
      parameters:
        - name: pet
          in: body
          required: true
          schema:
            $ref: '#/definitions/newPet'
      responses:
        default:
          description: error payload
  /pets/{id}:
    get:
      parameters:
        - name: id
          in: path
          required: true
          type: integer
          format: int64
      responses:
        default:
          description: error payload
    delete:
      parameters:
        - name: id
          in: path
          required: true
          type: integer
          format: int64
      responses:
        default:
          description: error payload
definitions:
  pet:
    required:
      - id
      - name
    properties:
      id:
        type: integer
        format: int64
      name:
        type: string
      tag:
        type: string
  newPet:
      required:
        - name
      properties:
        id:
          type: integer
          format: int64
        name:
          type: string
        tag:
          type: string

The get method on path /pets takes an optional limit parameter of common type integer. The post method takes a newPet body parameter comprising of the primitive attributes id, name and tag, subsequently of common types long and string (twice). Of these, only the name attribute is mandatory. The get method on the path /pets/{id} takes the path parameter id of common type long and returns an array of pets consisting of the same attributes and primitive types as a newPet - but this time with both name and id being mandatory. This specification maps to the following managed Scala domain model code:

package example

package object yaml {

    import de.zalando.play.controllers.PlayPathBindables

    type PetsIdDeleteResponsesDefault = Null
    type NewPetTag = Option[String]
    type PetsIdDeleteId = Long
    type PetsGetLimit = Option[Int]
    type NewPetId = Option[Long]

    case class Pet(id: Long, name: String, tag: NewPetTag) 
    case class NewPet(name: String, id: NewPetId, tag: NewPetTag) 

    implicit val bindable_OptionIntQuery = PlayPathBindables.createOptionQueryBindable[Int]
}

We want to have test data generators that generate an arbitrary range of values for the model code shown above - composed from primitive, and sometimes optional, data definitions. The Play-Swagger plugin does this by generating two Scala objects: one for the Swagger API definition, and one for the API path parts. Each object contains generator factory methods for the defined data types, prefixed by create, which returns a generator function. A generator function takes a given integer count and returns a generated amount of test data for the data type it was created for.

Data types are composed from primitive types, Scala optional types, and possibly more complex types. Test data values for the primitive types are generated arbitrarily, employing the ScalaCheck org.scalacheck.Arbitrary.arbitrary[T] method (the type parameter, replaced with Scala's primitive type, on which the Swagger common type is mapped).

In the code shown below, starting with primitive leaf data values, the pet parameter's attribute id of common type long is arbitrarily generated from a scala.Long. Note that the id attribute is optional, though, for the newPet definition. As with the generated model, we created a NewPetIdGenerator value that takes an arbitrarily generated scala.Long id value and generates an option value from it, employing the ScalaCheck org.scalacheck.Gen.option[T]. This generator will generate test data values comprising of None and Some arbitrarily id value. It's probably best to let the Scala generator code speak for itself. Note how it composes according to the same structure as the Scala model code.

package example.yaml

import org.scalacheck.Gen
import org.scalacheck.Arbitrary
import play.api.libs.json.scalacheck.JsValueGenerators
import Arbitrary._

object Generators extends JsValueGenerators {

    def createNullGenerator = _generate(NullGenerator)
    def createNewPetTagGenerator = _generate(NewPetTagGenerator)
    def createLongGenerator = _generate(LongGenerator)
    def createPetsGetLimitGenerator = _generate(PetsGetLimitGenerator)
    def createNewPetIdGenerator = _generate(NewPetIdGenerator)

    def createPetGenerator = _generate(PetGenerator)
    def createNewPetGenerator = _generate(NewPetGenerator)

    def NullGenerator = arbitrary[Null]
    def NewPetTagGenerator = Gen.option(arbitrary[String])
    def LongGenerator = arbitrary[Long]
    def PetsGetLimitGenerator = Gen.option(arbitrary[Int])
    def NewPetIdGenerator = Gen.option(arbitrary[Long])

    def PetGenerator = for {
        id <- arbitrary[Long]
        name <- arbitrary[String]
        tag <- NewPetTagGenerator
    } yield Pet(id, name, tag)
    def NewPetGenerator = for {
        name <- arbitrary[String]
        id <- NewPetIdGenerator
        tag <- NewPetTagGenerator
    } yield NewPet(name, id, tag)

    def _generate[T](gen: Gen[T]) = (count: Int) => for (i <- 1 to count) yield gen.sample
}

A PetGenerator and NewPetGenerator are created and implemented by the plugin as a for comprehension that generates data values for each attribute, yielding an instance of a test pet. Other generators follow the same pattern but, if necessary, delegate to different child generators. From this we acquire a set of test data generators to implement our property-based testing.

Running the test is as simple as running a test set from sbt. Just type test from your sbt prompt.

Building a Play-Swagger Plugin

To build a plugin, do the following:

  • Clone the repository to your local filesystem
  • Run sbt +publishLocal in the Play-Swagger directory. This will publish the plugin into your local ivy repository

To use the plugin in a plain Play project:

  • Create a new Play-Swagger project using activator template, for example: activator new hello-world play-swagger-service
  • Take a look at the project/plugins.sbt of the generated project and add required plugins and resolvers to the project/plugins.sbt of your Play project
  • Do the same for build.sbt
  • Put a Swagger specification with a .yaml or .json extension into the conf directory
  • Add a specification link (->) to the play's routes file

Plugin Architecture

Ths Play-Swagger plugin has a three-tier architecture:

  • specification - this tier is responsible for finding and parsing a specification and converting it into the raw AST format
  • normalisation - this tier performs a couple of optimisations on the AST including type deduplication, flattening and parameter dereferencing
  • generation - a final step including transformation of the AST into the source-code related data and generation of source code from it

The separation of the specification and generation tiers allows for plugging in different specification standards and generating source code for different frameworks.

Plugin Project Structure

There are a couple of sub-projects:

  • swagger-model - A standalone Scala Swagger model and a Jackson parser for it. Can be used by another projects
  • api - This is the project that's automatically added to the runtime classpath of any projects that use this plugin.
  • swagger-parser - A converter of the Swagger model to the internal AST of the plugin
  • api-first-core - This is a core of the plugin with minimal functionality. It includes defining an AST structure and some transformations on AST.
  • play-scala-generator - The standalone generator for transforming an AST into the skeleton of Play-Scala application.
  • plugin - A coupble of sbt plugins, one for each tier:
    • ApiFirstSwaggerParser - a plugin wrapping Swagger parsing part
    • ApiFirstCore - a wrapper for AST-related functionality
    • ApiFirstPlayScalaCodeGenerator - a wrapper for the Play-Scala generator

Because of the modular plugin architecture, all modules must be enabled separatly in sbt's build.sbt. It is also necessary to configure which parser(s) must be used by the plugin, like that:

lazy val root = (project in file(".")).enablePlugins(PlayScala, ApiFirstCore, ApiFirstPlayScalaCodeGenerator, ApiFirstSwaggerParser)

apiFirstParsers := Seq(ApiFirstSwaggerParser.swaggerSpec2Ast.value).flatten

Please take a look at activator template's configuration for complete example.

Custom Templates For Code Generation

The PlayScala generator supports custom templates. In order to override default template for some of the components, please provide your custom template named in accordance to the following list:

* `play_scala_test.mustache` - for unit tests
* `play_validation.mustache` - for validators 
* `generators.mustache` - for test data generators
* `model.mustache` - for model classes and query and path bindables
* `play_scala_controller_base.mustache` - for play controller bases 
* `play_scala_controller_security.mustache` - for security adapters used by controller bases
* `play_scala_form_parser.mustache` - for form parsers used by the controller bases
* `play_scala_controller.mustache` - for play controller skeletons supposed to be augmented by the programmer
* `play_scala_response_writers.mustache` - for custom serializers to be augmented by the programmer
* `play_scala_security_extractors.mustache` - for custom security extractors to be augmented by the programmer 

Please be aware that generated artifacts need to preserve some specific shape in order to be compiled together without errors.

The location where custom templates reside needs to be configured by overriding the plugin setting playScalaCustomTemplateLocation.

For example following configuration will set this place to be conf/templates folder of the project:

playScalaCustomTemplateLocation := Some(((resourceDirectory in Compile) / "templates").value)

Plugin Developing

sbt doesn't allow sub-projects to depend on each other as sbt plugins. To test an sbt plugin, you need a separate project. This project is swagger-tester. To test your changes as you're developing the plugin, cd into this directory, and run sbt. This project uses an sbt ProjectRef to the sbt plugin, which means you don't need to publishLocal the plugin after each change. Just run reload in the sbt console, and it will pick up your changes.

The play-swagger plugin provides a couple of commands useful for development:

  • apiFirstPrintDenotations - outputs a common names of different parts of the AST as they are intended to be used in generated Scala code
  • apiFirstPrintRawAstTypes - outputs all type definitions as they read from the specification before type optimisations
  • apiFirstPrintRawAstParameters - outputs all parameters definitions before type optimisations
  • apiFirstPrintFlatAstTypes - outputs type definitions after type optimisations
  • apiFirstPrintFlatAstParameters - outputs parameter definitions after type optimisations

Plugin Testing

We're using the sbt scripted framework for testing. You can find the tests in plugin/src/sbt-test, and run them by running scripted in the sbt console.

Code quality

There are some quality checks embedded into the build script:

  • the source code is (re)formatted using scalariform each time it is compiled (currently deactivated).
  • scalastyle sbt command shall be used to perform code style checks before putting changes into the repository.
  • lint:compile sbt command shall be used to perform static code analysis before putting changes into the repository.
  • code coverage for api and compiler modules can be executed by issuing sbt clean coverage test command for these projects. Coverage statistics can be generated using coverageReport sbt command.