json4s object extraction with extra data

Question

I'm using spray with json4s, and I've got the implementation below to handle put requests for updating objects... My problem with it, is that I first extract an instance of SomeObject from the json, but being a RESTful api, I want the ID to be specified in the URL. So then I must somehow create another instance of SomeObject that is indexed with the ID... to do this, I'm using a constructor like SomeObject(id: Long, obj: SomeObject). It works well enough, but the implementation is ugly and it feels inefficient. What can I do so I can somehow stick the ID in there so that I'm only creating one instance of SomeObject?

class ApplicationRouter extends BaseRouter {
  val routes =
    pathPrefix("some-path") {
      path("destination-resource" \ IntNumber) { id =>
        entity(as[JObject]) { rawData =>
          val extractedObject = rawData.camelizeKeys.extract[SomeObject]
          val extractedObjectWithId = SomeObject(id, extractedObject)
          handleRequest(extractedObjectWithId)
        }
      }
    }
}

case class SomeObject(id: Long, data: String, someValue: Double, someDate: DateTime) {
  def this(data: String, someValue: Double, someDate: DateTime) = this(0, data, someValue, someDate)
  def this(id: Long, obj: SomeObject) = this(id, obj.data, obj.someValue, obj.someDate)
}

score 1 · Accepted Answer · answered Oct 24 '14 at 16:23

I figured out a solution to this after digging around through the docs for awhile:

class ApplicationRouter extends BaseRouter {
  val routes =
    pathPrefix("some-path") {
      path("destination-resource" \ IntNumber) { id =>
        entity(as[JObject]) { rawData =>
          val extractedObject = rawData.camelizeKeys.merge { 
              ("id", id)
          }.extract[SomeObject]
          handleRequest(extractedObject)
        }
      }
    }
}

score 0 · Answer 2 · answered Oct 20 '14 at 10:58

Since id field is not set on all instances, it means it is optional so use Option type to indicate it. Define your case class with id: Option[Long] field. This make json parser to skip id field when it is absent but allow you to assign a value when you have.

case class SomeObject(id: Option[Long], data: String, someValue: Double, someDate: DateTime)

class ApplicationRouter extends BaseRouter {
  val routes =
    pathPrefix("some-path") {
      path("destination-resource" \ IntNumber) { id =>
        entity(as[JObject]) { rawData =>
          val extractedObject = rawData.camelizeKeys.extract[SomeObject]
          val extractedObjectWithId = extractedObject.copy(id = Some(id))
          handleRequest(extractedObjectWithId)
        }
      }
    }
}

And don't worry about performance impacts of creating new objects. It may affect performance much less than you thought. You should measure performance before improving it.

If I am not mistaken, the ID is not optional, but is not present in the JSON value, because it comes from a different source: The URL. In other words, making `id` optional would be semantically wrong. — Kulu Limpa, Oct 21 '14 at 12:48

score -1 · Answer 3 · answered Oct 17 '14 at 23:17

I do not know about efficiency, but you can make your code "less ugly" by defining a SomeObjectBuilder, to which you extract your JSON value.

case class SomeObjectBuilder(data: String, someValue: Double, someDate: DateTime) {
  def setId(id: Long) = SomeObject(id, data, someValue, someDate)
}

case class SomeObject(id: Long, data: String, someValue: Double, someDate: DateTime)

With the extraction:

class ApplicationRouter extends BaseRouter {
  val routes =
    pathPrefix("some-path") {
      path("destination-resource" \ IntNumber) { id =>
        entity(as[JObject]) { rawData =>
          val extractedObject = rawData.camelizeKeys.extract[SomeObjectBuilder]
          val extractedObjectWithId = extractedObject.setId(id)
          handleRequest(extractedObjectWithId)
        }
      }
    }
}

This way, you are not using a default id set to zero, which is, if I understand correctly, never correct. The only reason you're setting it to zero is that the value is not known by the extractor, so, using the builder, you make a partial instantiation explicit.

I can't really think of a better solution than this (though I wouldn't use setId for the function name there), but it still doesn't feel right. I was hoping there was some way for scala to extract a case class in to a tuple, and then for scala to pass the parameters of that tuple as parameters of the function. so it would come out like `SomeObject(id, someObject.toTuple.asParams)` Then scala would use a compiler macro to extract the parameters so it actually reads as something like `SomeObject(id, someObject.toTuple._1, someObject.toTuple._2, someObject.toTuple._3)` oh well. — JBarber, Oct 18 '14 at 18:54
Since your ID is known prior to extracting, you could also define a custom deserializer. This is done by extending `CustomSerializer`, which takes a partial function describing deserialization. The "tuple" you mention sounds much like providing the `id` as a closure to this partial function. However, in my opinion, implementing a custom serializer makes the code more complex than setting the `id` manually. — Kulu Limpa, Oct 19 '14 at 03:16

json4s object extraction with extra data

3 Answers3