Scala macros/paradise case class apply method

Question

I'm working on Scala 2.12.17.

Let's say I've a bunch of case classes :

case class TestOne(one: String)
case class TestTwo(one: String, two: String)
case class TestThree(one: String, two: String, three: String)

I also have these types :

  trait Data{
    val a: Int
  }
  
  case class DoubleInt(a: Int, b: Int) extends Data
  case class SingleInt(a: Int) extends Data

And this function which converts Data objects to String :

  def loadData(input: Data): String = {
    input.a.toString
  }

What I'm looking forward is to pass Data object(s) to my case classe's apply method, then the apply method would use loadData function in order to convert each passed Data object into a String to make an instance of my case class. E.g :

val dataOne: Data = SingleInt(1)
val dataTwo: Data = DoubleInt(1, 2)

val testOne = TestOne(dataOne)
val testTwo = TestTwo(dataOne, dataTwo)
val testThree = TestOne(dataOne, dataTwo, dataOne)

Basically, TestOne apply method would be :

def apply(one: Data): TestOne = {
  new TestOne(loadData(one))
}

TestTwo apply method would be :

def apply(one: Data, two: Data): TestTwo= {
  new TestTwo(loadData(one), loadData(two))
}

Is there any way to programatically generate those apply methods at compile time ?

I thought that macros or paradise annotations would be useful for this use case, but I'm too unexperienced with these topics to even know where to start :/

Dmytro Mitin · Accepted Answer · 2023-03-15T15:11:27.830

Should

val testThree = TestOne(dataOne, dataTwo, dataOne)

be val testThree = TestThree(dataOne, dataTwo, dataOne)?

So you'd like to replace

val testOne = TestOne(loadData(dataOne))
val testTwo = TestTwo(loadData(dataOne), loadData(dataTwo))
val testThree = TestThree(loadData(dataOne), loadData(dataTwo), loadData(dataOne))

with just

val testOne = TestOne(dataOne)
val testTwo = TestTwo(dataOne, dataTwo)
val testThree = TestThree(dataOne, dataTwo, dataOne)

Mapping over a tuple or case class is a standard task for example for Shapeless

// libraryDependencies += "com.chuusai" %% "shapeless" % "2.3.10"
import shapeless.poly.->
import shapeless.syntax.std.tuple._

object loadDataPoly extends (Data -> String)(loadData)

val testOne = TestOne(loadData(dataOne))
val testTwo = TestTwo.tupled((dataOne, dataTwo).map(loadDataPoly))
val testThree = TestThree.tupled((dataOne, dataTwo, dataOne).map(loadDataPoly))

If you want to hide load at all you can define a generic method

import shapeless.{Generic, HList}
import shapeless.ops.traversable.FromTraversable

def make[A <: Product] = new PartiallyAppliedMake[A]

class PartiallyAppliedMake[A <: Product] {
  def apply[L <: HList](data: Data*)(implicit
    generic: Generic.Aux[A, L],
    fromTraversable: FromTraversable[L]
  ): A = generic.from(fromTraversable(data.map(loadData)).get)
}

val testOne = make[TestOne](dataOne)
val testTwo = make[TestTwo](dataOne, dataTwo)
val testThree = make[TestThree](dataOne, dataTwo, dataOne)
val testThree_ = make[TestThree](dataOne, dataTwo, dataOne, dataOne) // fails at runtime

If you want make to fail at compile time if the number of arguments is incorrect then the definition is a little more complicated

import shapeless.{Generic, HList, Nat}
import shapeless.ops.hlist.{Mapper, Length, Fill}
import shapeless.ops.function.FnFromProduct

def make[TestClass <: Product] = new PartillyAppliedMake[TestClass]

class PartillyAppliedMake[TestClass <: Product] {
  def apply[StringHList <: HList, N <: Nat, DataHList <: HList]()(implicit
    generic: Generic.Aux[TestClass, StringHList],
    length: Length.Aux[StringHList, N],
    fill: Fill.Aux[N, Data, DataHList],
    mapper: Mapper.Aux[loadDataPoly.type, DataHList, StringHList],
    fnFromProduct: FnFromProduct[DataHList => TestClass]
  ): fnFromProduct.Out = 
    fnFromProduct((l: DataHList) => generic.from(mapper(l)))
}

val testOne = make[TestOne]().apply(dataOne)
val testTwo = make[TestTwo]().apply(dataOne, dataTwo)
val testThree = make[TestThree]().apply(dataOne, dataTwo, dataOne)
val testThree_ = make[TestThree]().apply(dataOne, dataTwo, dataOne, dataOne) // fails at compile time

Or you can define a def macro

// libraryDependencies += scalaOrganization.value % "scala-reflect" % scalaVersion.value
import scala.language.experimental.macros
import scala.reflect.macros.blackbox

def make[A](data: Data*): A = macro makeImpl[A]

def makeImpl[A: c.WeakTypeTag](c: blackbox.Context)(data: c.Tree*): c.Tree = {
  import c.universe._
  val A = weakTypeOf[A]
  val strs = data.map(t => q"loadData($t)")
  q"new $A(..$strs)"
}

// in a different subproject

val testOne = make[TestOne](dataOne) // TestOne(1)
val testTwo = make[TestTwo](dataOne, dataTwo) // TestTwo(1,1)
val testThree = make[TestThree](dataOne, dataTwo, dataOne) // TestThree(1,1,1)
val testThree_ = make[TestThree](dataOne, dataTwo, dataOne, dataOne) // doesn't compile: too many arguments (found 4, expected 3) ...

    // scalacOptions += "-Ymacro-debug-lite"
//scalac: new App.TestOne(loadData(App.this.dataOne))
//scalac: new App.TestTwo(loadData(App.this.dataOne), loadData(App.this.dataTwo))
//scalac: new App.TestThree(loadData(App.this.dataOne), loadData(App.this.dataTwo), loadData(App.this.dataOne))

But if you really want to generate apply methods in companion objects you can define macro annotaion (settings: Auto-Generate Companion Object for Case Class in Scala)

// addCompilerPlugin("org.scalamacros" % "paradise" % "2.1.1" cross CrossVersion.full) // Scala 2.12
import scala.annotation.{StaticAnnotation, compileTimeOnly}
import scala.language.experimental.macros
import scala.reflect.macros.blackbox

@compileTimeOnly("enable macro annotations")
class generateApply extends StaticAnnotation {
  def macroTransform(annottees: Any*): Any = macro GenerateApplyMacro.macroTransformImpl
}
object GenerateApplyMacro {
  def macroTransformImpl(c: blackbox.Context)(annottees: c.Tree*): c.Tree = {
    import c.universe._

    def modify(cls: ClassDef, obj: ModuleDef): Tree = (cls, obj) match {
      case (
        q"$_ class $tpname[..$tparams] $_(...$paramss) extends { ..$_ } with ..$_ { $_ => ..$_ }",
        q"$mods object $tname extends { ..$earlydefns } with ..$parents { $self => ..$body }"
      ) =>
        val paramss1 = paramss.map(_.map {
          case q"$mods val $tname: $_ = $_" => q"$mods val $tname: Data"
        })
        val argss = paramss.map(_.map {
          case q"$_ val $tname: $_ = $_" => q"loadData($tname)"
        })
        val targs = tparams.map {
          case q"$_ type $tpname[..$_] = $tpt" => tq"$tpname"
        }
        q"""
          $cls

          $mods object $tname extends { ..$earlydefns } with ..$parents { $self =>
            def apply[..$tparams](...$paramss1): $tpname[..$targs] = {
              new $tpname[..$targs](...$argss)
            }
            ..$body
          }
        """
    }

    annottees match {
      case (cls: ClassDef) :: (obj: ModuleDef) :: Nil => modify(cls, obj)
      case (cls: ClassDef) :: Nil => modify(cls, q"object ${cls.name.toTermName}")
    }
  }
}

// in a different subproject

@generateApply case class TestOne(one: String)
@generateApply case class TestTwo(one: String, two: String)
@generateApply case class TestThree(one: String, two: String, three: String)

val testOne = TestOne(dataOne) // TestOne(1)
val testTwo = TestTwo(dataOne, dataTwo) // TestTwo(1,1)
val testThree = TestThree(dataOne, dataTwo, dataOne) // TestThree(1,1,1)

//scalac: {
//  case class TestOne extends scala.Product with scala.Serializable {
//    <caseaccessor> <paramaccessor> val one: String = _;
//    def <init>(one: String) = {
//      super.<init>();
//      ()
//    }
//  };
//  object TestOne extends scala.AnyRef {
//    def <init>() = {
//      super.<init>();
//      ()
//    };
//    def apply(one: Data): TestOne = new TestOne(loadData(one))
//  };
//  ()
//}

//scalac: {
//  case class TestTwo extends scala.Product with scala.Serializable {
//    <caseaccessor> <paramaccessor> val one: String = _;
//    <caseaccessor> <paramaccessor> val two: String = _;
//    def <init>(one: String, two: String) = {
//      super.<init>();
//      ()
//    }
//  };
//  object TestTwo extends scala.AnyRef {
//    def <init>() = {
//      super.<init>();
//      ()
//    };
//    def apply(one: Data, two: Data): TestTwo = new TestTwo(loadData(one), loadData(two))
//  };
//  ()
//}

//scalac: {
//  case class TestThree extends scala.Product with scala.Serializable {
//    <caseaccessor> <paramaccessor> val one: String = _;
//    <caseaccessor> <paramaccessor> val two: String = _;
//    <caseaccessor> <paramaccessor> val three: String = _;
//    def <init>(one: String, two: String, three: String) = {
//      super.<init>();
//      ()
//    }
//  };
//  object TestThree extends scala.AnyRef {
//    def <init>() = {
//      super.<init>();
//      ()
//    };
//    def apply(one: Data, two: Data, three: Data): TestThree = new TestThree(loadData(one), loadData(two), loadData(three))
//  };
//  ()
//}

Is it normal that I'm having a Warning on `val testOne = TestOne(dataOne) // TestOne(1)` in IntelliJ ? It says that the required type needs to be `String` — Fragan, Mar 15 '23 at 18:01
@Fragan IntelliJ warnings don't matter. You can check with `sbt clean compile`. — Dmytro Mitin, Mar 15 '23 at 18:10

Scala macros/paradise case class apply method

1 Answers1