rallyhealth / play-json-ops

Common Play Json implicits and helper methods

Version Matrix

Build Status codecov

Play Json Ops

Augments the Play Json library with some helpful implicits and tools for:

  • Creating formats for traits and abstract classes
  • Safely printing error messages with redacted sensitive data using implicit transformations
  • Formats for tuples (up to arity-10) as JsArray
  • Formats for scala.concurrent.Duration
  • Safe formats for Map via KeyReads and KeyWrites
  • Format builder for empty collections
  • UTCFormats for org.joda.time.DateTime
  • ScalaCheck generators for JsValue, JsArray, and JsObject

Versions

4.X Branch

play version scala versions scalatest version artifact name bintray
2.8.x 2.13.1, 2.12.6, 2.11.12 play28-json-ops-v4 Download
2.8.x 2.13.1, 2.12.6, 2.11.12 3.1.x play28-json-tests-sc14 Download
2.7.x 2.13.1, 2.12.6, 2.11.12 play27-json-ops-v4 Download
2.7.x 2.13.1, 2.12.6, 2.11.12 3.1.x play27-json-tests-sc14 Download
2.6.x 2.12.6, 2.11.12 play26-json-ops-v4 Download
2.6.x 2.12.6, 2.11.12 3.0.x play26-json-tests-sc13 Download
2.5.x 2.11.12 play25-json-ops-v4 Download
2.5.x 2.11.12 3.1.x play25-json-tests-sc14 Download
2.5.x 2.11.12 3.0.x play25-json-tests-sc13 Download
2.5.x 2.11.12 2.2.x play25-json-tests-sc12 Download

3.X Branch

play version scala versions scalatest version artifact name bintray
2.7.x 2.13.1, 2.12.6, 2.11.12 play27-json-ops Download
2.7.x 2.13.1, 2.12.6, 2.11.12 3.1.x play27-json-tests-sc14 Download
2.6.x 2.12.6, 2.11.12 play26-json-ops Download
2.6.x 2.12.6, 2.11.12 3.0.x play26-json-tests-sc13 Download
2.5.x 2.11.12 play25-json-ops Download
2.5.x 2.11.12 3.1.x play25-json-tests-sc14 Download
2.5.x 2.11.12 3.0.x play25-json-tests-sc13 Download
2.5.x 2.11.12 2.2.x play25-json-tests-sc12 Download

2.X Branch

play version scala versions scalatest version artifact name artifact version
2.5.x 2.11.12 play25-json-ops 2.1.1
2.5.x 2.11.12 3.0.x play25-json-tests 2.1.1
2.5.x 2.11.12 2.2.x play25-json-tests-sc12 2.1.1
2.3.x 2.11.12 play23-json-ops 2.1.1
2.3.x 2.11.12 3.0.x play23-json-tests 2.1.1
2.3.x 2.11.12 2.2.x play23-json-tests-sc12 2.1.1

Getting Started

Pretty much all of these tools become available when you import play.api.libs.json.ops._

Dependencies

  • scalacheck-ops: for the ability to convert ScalaCheck Gen into an Iterator

Features

Implicits

By importing play.api.libs.json.ops._, you get access to:

  • PlayJsonMacros.nullableReads macro that will read null as [] for all container fields of a case class
  • Reads, Format, and OFormat extension methods to recover from exceptions
  • Many extension methods for the play.api.libs.json.Json
    • Format.of[A], OFormat.of[A], and OWrites.of[A] for summoning formats the same as Reads.of[A] and Writes.of[A]
    • Format.asEither[A, B] for reading and writing an either value based on some condition
    • Format.asString[A] for reading and writing a wrapper type as a string
    • Format.pure for reading and writing a constant value
    • Format.empty for reading or writing an empty collection
    • In Play 2.3, the Json.format and Json.writes macros would return Format and Writes instead of OFormat and OWrites, even though the macros would only produce these types. The play-json-ops for Play 2.3 provides a Json.oformat and Json.owrites which uses the underlying Play Json macros, but it casts the results.
  • Reads and Writes implicits for tuple types (encoded as a JsArray)
  • The JsValue extension method .asOrThrow[A] which throws a better exception that .as[A]
  • And handy syntax for the features listed below

Tolerant Container Reads Macro

Extending the TolerantContainerFormats trait or importing from its companion object will give you the ability to call .readNullableContainer on a Reads instance. This will allow you to parse null fields as empty collections.

You can also use PlayJsonMacros.nullableReads to create a Reads for a case class that will accept either null or missing field values for any container fields (Seq, Set, Map, etc) using the same method.

case class Example(values: Seq[Int])
object Example extends TolerantContainerFormats {

  val nonMacroExample: Reads[Seq[Int]] = (__ \ "values").readNullableContainer[Seq, Int]
  assert(Json.parse("null").as(nonMacroExample) == JsSuccess(Seq()))
  assert(Json.parse("[]").as[Example] == JsSuccess(Seq()))
  assert(Json.parse("[1]").as[Example] == JsSuccess(Seq(1)))

  val macroExample: Reads[Example] = PlayJsonMacros.nullableReads[Example]
  assert(Json.parse("{}").as(macroExample) == JsSuccess(Example(Seq())))
  assert(Json.parse("""{"values":null}""").as(macroExample) == JsSuccess(Example(Seq())))
  assert(Json.parse("""{"values":[]}""").as(macroExample) == JsSuccess(Example(Seq())))
  assert(Json.parse("""{"values":[1]}""").as(macroExample) == JsSuccess(Example(Seq(1))))
}

Reads Recovery Methods

You can call .recoverJsError, .recoverTotal, or .recoverWith on a Reads, Format, or OFormat instance. These methods allow you to recover from exceptions thrown during the reading process into an appropriate JsResult.

object ReadsRecoveryExamples {

  // converts all exceptions into a JsError with the exception captured as an argument in the JsonValidationError
  val readIntAsString = Reads.of[String].map(_.toInt).recoverJsError
  assert(readIntAsString.reads("not a number").isError) // no exception thrown

  // converts only the matched exceptions to JsResults, all others continue to throw
  val invertReader = Reads.of[String].map(1 / _.toDouble).recoverWith {
    case _: ArithmeticException => JsSuccess(Double.MaxValue) 
  }
  invertReader.reads("not a number") // throws NumberFormatException
  assert(invertReader.reads("0") == JsSuccess(Double.MaxValue)) // handles ArithmeticException

  // converts all exceptions into some value of the right type
  val readAbsValueOrSentinel = Reads.of[String].map(_.toInt.abs).recoverTotal(_ => -1)
  assert(readAbsValueOrSentinel.reads("not a number") == JsSuccess(-1))

  // these can be combined, of course
  val safeInvertReader = invertReader.recoverJsError
  assert(safeInvertReader.reads("not a number").isError) // no exception thrown
}

Automatic Automated Tests

To get free test coverage, just extend PlayJsonFormatSpec[T] where T is a serializable type that you would like to create a suite of tests for. All it requires is a ScalaCheck generator of the same type or a sequence of examples.

This will use ScalaTest to create the test cases, however it will work just as well with Specs2

case class Example(value: String)

object Example {
  implicit val format = Json.format[Example]
}

object ExampleGenerators {
  implicit def arbExample(implicit arbString: Arbitrary[String]): Arbitrary[Example] =
    Arbitrary(arbString.map(Example(_)))
}

import ExampleGenerators._

// Free unit tests for serializing and deserializing Example values
// Also works with implicit Shrink[Example]
class ExampleFormatSpec extends PlayJsonFormatSpec[Example]

Creating Formats for Traits and Abstract Classes

The following example shows how you can create a Format for the Generic trait using Json.formatAbstract. This method requires an implicit TypeKeyExtractor[Generic], which is used to pull a "key" value from some field in the json / model. This key value is then matched on by a provided partial function from key to format: Any => OFormat[_ <: Generic].

The pattern works as follows:

  1. Create the formats of each of the specific formats using Json.formatWithType and the Json.format macro.

    This will append the key field (even if it isn't in the case class constructor args) to the output json.

  2. Create an implicit TypeKeyExtractor for the generic trait or abstract class on the companion object of that class.

    This is required for the Json.formatWithType to work properly and avoids repeating unnecessary boilerplate on each of the specific serializers to write out the key or the generic serializer to read the key.

  3. Finally, define an implicit Format for your generic trait or abstract class using Json.formatAbstract by providing a partial function from the extracted key (from #2) to the specific serializer (from #1). Any unmatched keys will throw an exception.

import play.api.libs.json._
import play.api.libs.json.ops._

sealed trait Generic {
  def key: String
}

object Generic {

  implicit val extractor: TypeKeyExtractor[Generic] =
    Json.extractTypeKey[Generic].usingKeyField(_.key, __ \ "kind")

  implicit val format: OFormat[Generic] = Json.formatAbstract[Generic] {
    case SpecificA.key => OFormat.of[SpecificA]
    case SpecificB.key => OFormat.of[SpecificB]
  }
}

case class SpecificA(value: String) extends Generic {

  override def key: String = SpecificA.key
}

object SpecificA {
  final val key = "A"

  // NOTE: You will need to use Json.oformat for Play 2.3.x
  implicit val format: OFormat[SpecificA] = Json.formatWithTypeKeyOf[Generic].addedTo(Json.format[SpecificA])
}

case class SpecificB(value: String) extends Generic {

  override def key: String = SpecificB.key
}

object SpecificB {
  final val key = "B"

  implicit val format: OFormat[SpecificB] = Json.formatWithTypeKeyOf[Generic].addedTo(Json.format[SpecificB])
}

case object SpecificC extends Generic {
  final val key = "C"

  implicit val format: OFormat[this.type] = OFormat.pure(this, Generic.extractor.writeKeyToJson(this))
}

Duration Formats

You can add implicit Json serializers by importing DurationFormat.string or DurationFormat.array depending on the format you want.

You can also extend ArrayDurationFormat or StringDurationFormat for the same effect, but it requires that you also extend an ImplicitDurationReads. A good default is to extend ForgivingDurationReads as this will read either format.

Ok, now how the formats look in Json:

  • ArrayDurationFormat

    [1, "seconds"]
  • StringDurationFormat

    "1 second"

ScalaCheck JsValue Generators

ScalaCheck is a very simple and powerful library for property-based testing.

Fun fact: It is the only library dependency of the Scala compiler

Ok, so assuming you are already familiar with ScalaCheck now... Let's say you want to generate arbitrary JsValues or JsObjects. All you have to do is extend JsValueGenerators in your test class and voila!

By default the maximum depth of the JsValue trees is set to JsValueGenerators.maxDepth and the maximum number of fields for JsObject and values for JsArray is set to JsValueGenerators.maxWidth. You can override this in local scope by providing an implicit Depth or Width type value:

implicit val maxDepth: Depth = 4  

forAll() { (json: JsValue) =>
  // ...
}

or passing the values explicitly:

forAll(genJsValue(maxDepth = 4, maxWidth = 12)) { (json: JsValue) =>
  // ...
}

Note: I encountered a compiler bug when overriding implicits in a local scope where the compiler would NOT throw the normal "ambiguous implicit values" exception and instead use the depth defined in the outer scope. Just be sure not to define ambiguous implicit Depth and Width values and everything works great.