Scala Automapper for Scala 2.12

I've just released version 0.4.0 of scala-automapper.

It brings support for Scala 2.12 and some nice API improvements.

The main use case for the library is automatic mapping between DTOs in a layered architecture.

It can also be quite handy in eventsourced or CQRS application, where commands usually need to be mapped one to one to events, and events need to be applied to some state.

Keep reading for some example usage.

Let's use the following classes for a very simple example:

case class SourceClass(label: String, value: Int)
case class TargetClass(label: String, value: Int)

To map a source class instance to the target class use any of the following ways:

import com.bfil.automapper._

val source = SourceClass("label", 10)
val target = automap(source).to[TargetClass]

Using implicit mappings

Implicit mappings can be defined separately and then used to map case classes

import com.bfil.automapper._

val source = SourceClass("label", 10)

trait MyMappings {
	implicit val mapping1 = generateMapping[SourceClass, TargetClass]
	implicit val mapping2 = generateMapping[SourceClass, AnotherClass]
}

object Example extends MyMappings {
	val target1 = automap(source).to[TargetClass]
	val target2 = automap(source).to[AnotherClass]
}

This example triggers the macro to generate the Mapping into the MyMappings trait, while the previous example used an implicit conversion to automatically generate the implicit mapping on the fly.

There's no real difference, obviously the first one is less verbose, but we will take a look at how to generate more complex mappings that require the mappings to be generated separately.

If some of the fields cannot be mapped automatically a compilation error will occur notifying the missing fields. In this case we can fill out the blanks by using dynamic mappings.

Dynamic mappings

It is pretty common to want to rename a field, or to have a calculated field into the target class that depend on the source class or other variables.

A dynamic mapping can be used to be able to partially map case classes with custom logic.

Take a look at the following example:

case class SourceClass(label: String, field: String, values: Int)
case class TargetClass(label: String, renamedField: String, total: Int)

The label field can be automatically mapped, but not the other 2, here is how you can specify a dynamic mapping for those fields:

import com.bfil.automapper._

val source = SourceClass("label", "field", List(1, 2, 3))

def sum(values: List[Int]) = values.sum

val target = automap(source).dynamicallyTo[TargetClass](
  renamedField = source.field, total = sum(values)
)

The example is unnecessarily complex just to demonstrate that it's possible to write any type of custom logic for the dynamic mapping (or at least I haven't found other issues so far).

Note that we didn't have to provide a value for the label field, since it could be automatically mapped.

Mapping rules

To fully understand how the mapping takes place here are some basic rules that are applied by the macro when generating the mapping:

  1. The dynamic mapping takes precedence over everything else
  2. Option fields will be filled in with a value of None if the source class does not contain the field
  3. Iterable and Map fields will be filled in with an empty Iterable / Map if the source class does not contain the field
  4. If the target class has a field with a default value it will be used if the source class does not contain the field
  5. Due to how the mapping is generated default values for Option / Iterable / Map fields will not be considered and a None or empty value will be used into the target class instead

Generated code

To give some insight on how the macro generated code looks like, here are some examples taken from the tests.

Here is our example source class:

case class SourceClass(
  field: String,
  data: SourceData,
  list: List[Int],
  typedList: List[SourceData],
  optional: Option[String],
  typedOptional: Option[SourceData],
  map: Map[String, Int],
  typedMap: Map[String, SourceData],
  level1: SourceLevel1)

case class SourceData(label: String, value: Int)
case class SourceLevel1(level2: Option[SourceLevel2])
case class SourceLevel2(treasure: String)

Without dynamic mapping

The code without dynamic mapping looks pretty much as it would look like if the mapping was created manually.

This is how the target class looks like, basically it's just a mirror of the source class:

case class TargetClass(
  field: String,
  data: TargetData,
  list: List[Int],
  typedList: List[TargetData],
  optional: Option[String],
  typedOptional: Option[TargetData],
  map: Map[String, Int],
  typedMap: Map[String, TargetData],
  level1: TargetLevel1)

case class TargetData(label: String, value: Int)
case class TargetLevel1(level2: Option[TargetLevel2])
case class TargetLevel2(treasure: String)

And here is the mapping generated by the macro:

{
  import com.bfil.automapper.Mapping;
  {
    final class $anon extends Mapping[SourceClass, TargetClass] {
      def map(a: SourceClass): TargetClass = TargetClass(
        field = a.field,
        data = TargetData(label = a.data.label, value = a.data.value),
        list = a.list,
        typedList = a.typedList.map(((a) => TargetData(label = a.label, value = a.value))),
        optional = a.optional,
        typedOptional = a.typedOptional.map(((a) => TargetData(label = a.label, value = a.value))),
        map = a.map,
        typedMap = a.typedMap.mapValues(((a) => TargetData(label = a.label, value = a.value))),
        level1 = TargetLevel1(level2 = a.level1.level2.map(((a) => TargetLevel2(treasure = a.treasure)))))
    };
    new $anon()
  }
}

With dynamic mapping

The code with dynamic mapping has the only overhead of having to use an instance of Dynamic, so it looks a little bit different.

This is how the target class looks like:

case class TargetWithDynamicMapping(renamedField: String, data: TargetData, total: Int)

Here is how the dynamic mapping looks like:

val values = source.list
def sum(values: List[Int]) = values.sum

automap(source).dynamicallyTo[TargetWithDynamicMapping](
  renamedField = source.field, total = sum(values)
)

And finally, here is the mapping generated by the macro:

{
  import com.bfil.automapper.Mapping;
  {
    final class $anon extends Mapping[SourceClass, TargetWithDynamicMapping] {
      def map(a: SourceClass): TargetWithDynamicMapping = {
        TargetWithDynamicMapping(
          renamedField = source.field,
          data = TargetData(label = a.data.label, value = a.data.value),
          total = sum(values)
        )
      }
    };
    new $anon()
  }
}

Pretty cool. Huh?