自动将案例类转换为无形的可扩展记录?

时间:2015-11-22 18:50:48

标签: scala shapeless

如果我有这两个案例类:

case class Address(street : String, zip : Int)
case class Person(name : String, address : Address)

和一个实例:

val person = Person("Jane", Address("street address", 12345))

有没有办法在无形状中自动将person转换为可扩展记录?

我对浅层和深层转换感兴趣。

浅层副本可能是:

'name ->> "Jane" :: 'address ->> Address("street address", 12345) :: HNil

在深度转换中,嵌套的case类也成为记录:

'name ->> "Jane" :: 'address ->> ('street ->> "street address" :: 'zip ->> 12345 :: HNil) :: HNil

我也有兴趣将记录转换回案例类。

1 个答案:

答案 0 :(得分:8)

假设我们有以下设置:

import shapeless._, shapeless.labelled.{ FieldType, field }

case class Address(street: String, zip: Int)
case class Person(name: String, address: Address)

val person = Person("Jane", Address("street address", 12345))

type ShallowPersonRec =
  FieldType[Witness.`'name`.T, String] ::
  FieldType[Witness.`'address`.T, Address] :: HNil

type DeepPersonRec =
  FieldType[Witness.`'name`.T, String] ::
  FieldType[
    Witness.`'address`.T,
    FieldType[Witness.`'street`.T, String] ::
    FieldType[Witness.`'zip`.T, Int] :: HNil
  ] :: HNil

Shapeless的LabelledGeneric直接支持浅层案例:

val shallow: ShallowPersonRec = LabelledGeneric[Person].to(person)

或者如果你想要一个通用的辅助方法:

def shallowRec[A](a: A)(implicit gen: LabelledGeneric[A]): gen.Repr = gen.to(a)

val shallow: ShallowPersonRec = shallowRec(person)

您可以使用from

返回
scala> val originalPerson = LabelledGeneric[Person].from(shallow)
originalPerson: Person = Person(Jane,Address(street address,12345))

深层的情况比较棘手,据我所知,没有方便的方法可以使用Shapeless提供的类型类和其他工具,但你可以调整this question的代码(现在是在{Shapeless}中test case做你想做的事。首先是类型类本身:

trait DeepRec[L] extends DepFn1[L] {
  type Out <: HList

  def fromRec(out: Out): L
}

然后是记录头部本身没有LabelledGeneric实例的情况的低优先级实例:

trait LowPriorityDeepRec {
  type Aux[L, Out0] = DeepRec[L] { type Out = Out0 }

  implicit def hconsDeepRec0[H, T <: HList](implicit
    tdr: Lazy[DeepRec[T]]
  ): Aux[H :: T, H :: tdr.value.Out] = new DeepRec[H :: T] {
    type Out = H :: tdr.value.Out    
    def apply(in: H :: T): H :: tdr.value.Out = in.head :: tdr.value(in.tail)
    def fromRec(out: H :: tdr.value.Out): H :: T =
      out.head :: tdr.value.fromRec(out.tail)
  }
}

然后是伴侣对象的其余部分:

object DeepRec extends LowPriorityDeepRec {
  def toRec[A, Repr <: HList](a: A)(implicit
    gen: LabelledGeneric.Aux[A, Repr],
    rdr: DeepRec[Repr]
  ): rdr.Out = rdr(gen.to(a))

  class ToCcPartiallyApplied[A, Repr](val gen: LabelledGeneric.Aux[A, Repr]) {
    type Repr = gen.Repr    
    def from[Out0, Out1](out: Out0)(implicit
      rdr: Aux[Repr, Out1],
      eqv: Out0 =:= Out1
    ): A = gen.from(rdr.fromRec(eqv(out)))
  }

  def to[A](implicit
    gen: LabelledGeneric[A]
  ): ToCcPartiallyApplied[A, gen.Repr] =
    new ToCcPartiallyApplied[A, gen.Repr](gen) 

  implicit val hnilDeepRec: Aux[HNil, HNil] = new DeepRec[HNil] {
    type Out = HNil    
    def apply(in: HNil): HNil = in
    def fromRec(out: HNil): HNil = out
  }

  implicit def hconsDeepRec1[K <: Symbol, V, Repr <: HList, T <: HList](implicit
    gen: LabelledGeneric.Aux[V, Repr],
    hdr: Lazy[DeepRec[Repr]],
    tdr: Lazy[DeepRec[T]]
  ): Aux[FieldType[K, V] :: T, FieldType[K, hdr.value.Out] :: tdr.value.Out] =
    new DeepRec[FieldType[K, V] :: T] {
      type Out = FieldType[K, hdr.value.Out] :: tdr.value.Out
      def apply(
        in: FieldType[K, V] :: T
      ): FieldType[K, hdr.value.Out] :: tdr.value.Out =
        field[K](hdr.value(gen.to(in.head))) :: tdr.value(in.tail)
      def fromRec(
        out: FieldType[K, hdr.value.Out] :: tdr.value.Out
      ): FieldType[K, V] :: T =
        field[K](gen.from(hdr.value.fromRec(out.head))) ::
          tdr.value.fromRec(out.tail)
    }
}

(请注意,必须同时定义DeepRec特征和对象以进行配对。)

这很麻烦,但确实有效:

scala> val deep: DeepPersonRec = DeepRec.toRec(person)
deep: DeepPersonRec = Jane :: (street address :: 12345 :: HNil) :: HNil

scala> val originalPerson = DeepRec.to[Person].from(deep)
originalPerson: Person = Person(Jane,Address(street address,12345))

转换回案例类的to / from语法是必要的,因为任何给定的记录都可能对应于非常多的潜在案例类,因此我们需要能够指定目标类型,由于Scala不支持部分应用的类型参数列表,我们必须将操作分解为两部分(其中一部分将明确指定其类型,而另一部分的类型参数将被推断)