19

I am trying to solve [this][1] question using Shapeless, in summary it's about converting a nested case class to Map[String,Any], here is the example:

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

val p = Person("Tom", Address("Jefferson st", 10000))

The goal is to convert p to following:

Map("name" -> "Tom", "address" -> Map("street" -> "Jefferson st", "zip" -> 10000))

I am trying to do it using Shapeless LabelledGeneric, here is what I have so far:

import shapeless._
import record._, syntax.singleton._
import ops.record._
import shapeless.ops.record._

def writer[T,A<:HList,H<:HList](t:T)
(implicit lGeneric:LabelledGeneric.Aux[T,A],
 kys:Keys.Aux[A,H],
 vls:Values[A]) = {
    val tGen = lGeneric.to(t)
    val keys = Keys[lGeneric.Repr].apply
    val values = Values[lGeneric.Repr].apply(tGen)
    println(keys)
    println(values)
  }

I am trying to have a recursive writer check each value and try to make Map for each element in value. The above code works fine but when I want to iterate over values with a sample Poly, using the following code I got these errors.

values.map(identity)
//or
tGen.map(identity)

Error:(75, 19) could not find implicit value for parameter mapper: shapeless.ops.hlist.FlatMapper[shapeless.poly.identity.type,vls.Out]
    values.flatMap(identity)
                  ^
Error:(75, 19) not enough arguments for method flatMap: (implicit mapper: shapeless.ops.hlist.FlatMapper[shapeless.poly.identity.type,vls.Out])mapper.Out.
Unspecified value parameter mapper.
    values.flatMap(identity)
                  ^

I don't know why I'm getting that error. I also would be happy to know if there is an easier way to do the whole thing using Shapeless. [1]: Scala macros for nested case classes to Map and other way around

2 Answers 2

26

Any time you want to perform an operation like flatMap on an HList whose type isn't statically known, you'll need to provide evidence (in the form of an implicit parameter) that the operation is actually available for that type. This is why the compiler is complaining about missing FlatMapper instances—it doesn't know how to flatMap(identity) over an arbitrary HList without them.

A cleaner way to accomplish this kind of thing would be to define a custom type class. Shapeless already provides a ToMap type class for records, and we can take it as a starting point, although it doesn't provide exactly what you're looking for (it doesn't work recursively on nested case classes).

We can write something like the following:

import shapeless._, labelled.FieldType, record._

trait ToMapRec[L <: HList] { def apply(l: L): Map[String, Any] }

Now we need to provide instances for three cases. The first case is the base case—the empty record—and it's handled by hnilToMapRec below.

The second case is the case where we know how to convert the tail of the record, and we know that the head is something that we can also recursively convert (hconsToMapRec0 here).

The final case is similar, but for heads that don't have ToMapRec instances (hconsToMapRec1). Note that we need to use a LowPriority trait to make sure that this instance is prioritized properly with respect to hconsToMapRec0—if we didn't, the two would have the same priority and we'd get errors about ambiguous instances.

trait LowPriorityToMapRec {
  implicit def hconsToMapRec1[K <: Symbol, V, T <: HList](implicit
    wit: Witness.Aux[K],
    tmrT: ToMapRec[T]
  ): ToMapRec[FieldType[K, V] :: T] = new ToMapRec[FieldType[K, V] :: T] {
    def apply(l: FieldType[K, V] :: T): Map[String, Any] =
      tmrT(l.tail) + (wit.value.name -> l.head)
  }
}

object ToMapRec extends LowPriorityToMapRec {
  implicit val hnilToMapRec: ToMapRec[HNil] = new ToMapRec[HNil] {
    def apply(l: HNil): Map[String, Any] = Map.empty
  }

  implicit def hconsToMapRec0[K <: Symbol, V, R <: HList, T <: HList](implicit
    wit: Witness.Aux[K],
    gen: LabelledGeneric.Aux[V, R],
    tmrH: ToMapRec[R],
    tmrT: ToMapRec[T]
  ): ToMapRec[FieldType[K, V] :: T] = new ToMapRec[FieldType[K, V] :: T] {
    def apply(l: FieldType[K, V] :: T): Map[String, Any] =
      tmrT(l.tail) + (wit.value.name -> tmrH(gen.to(l.head)))
  }
}

Lastly we provide some syntax for convenience:

implicit class ToMapRecOps[A](val a: A) extends AnyVal {
  def toMapRec[L <: HList](implicit
    gen: LabelledGeneric.Aux[A, L],
    tmr: ToMapRec[L]
  ): Map[String, Any] = tmr(gen.to(a))
}

And then we can demonstrate that it works:

scala> p.toMapRec
res0: Map[String,Any] = Map(address -> Map(zip -> 10000, street -> Jefferson st), name -> Tom)

Note that this won't work for types where the nested case classes are in a list, tuple, etc., but you could extend it to those cases pretty straightforwardly.

Sign up to request clarification or add additional context in comments.

14 Comments

Thanks for awesome magics! One question, how would be the other way, from Map[String,Any] to case class, is there a good way to do it because we have lost all the types?
@Omid It's possible if you know the target type and are willing to receive the result in an Option (or some other type that can represent failure).
@Omid Great! It's probably worth a new question, though. :)
@Omid, oh no, sorry—you'll need a ToMapRec for the generic representation of A, not A itself. Probably worth a new question to avoid drawing out the comments further.
@Omid Instead of returning ToMapRec[FieldType[K, V] :: T] you'd have a method returning ToMapRec[FieldType[K, List[V]] :: T], etc. Probably worth a new question.
|
3

I have a problem with an approach provided by Travis Brown.
Some of nesting case classes are not converted to Map https://scalafiddle.io/sf/cia2jTa/0.

The answer was found here.
To correct the solution just wrap ToMapRec[T] in implicit parameters to Lazy[ToMapRec[T]]. Corrected fiddle https://scalafiddle.io/sf/cia2jTa/1

Comments

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.