Reduce compilation times with the @reader annotation about grafter HOT 10 CLOSED

Wow, that's pretty significant. I'll see if can figure something out.

from grafter.

I also tested the @readers annotation on our ApplicationConfig and replaced it with 14 implicit definitions. I observed a further drop of 10 seconds.

from grafter.

My (admittedly simple) test application doesn't seem to have such dramatic differences. Which versions of Scala/SBT are you using?

from grafter.

Scala 2.11.8 / Sbt 0.13.13

from grafter.

It looks like the @readers annotation makes compilation more performant but having @reader annotation on components degrades the compilation times. The likely explanation is that macro expansion is costly but it is also possible that unnecessary work is done.

The next step would be to compare the AST generated with the @reader macro and with the manual implicit def reader definition.

from grafter.

The AST for the annotated case class in ReaderMacroTest prints as

List(case class C extends scala.Product with scala.Serializable {
  def <init>() = {
    super.<init>();
    ()
  }
}, object C extends scala.AnyRef {
  def <init>() = {
    super.<init>();
    ()
  };
  import org.zalando.grafter.GenericReader._;
  implicit def reader: cats.data.Reader[ApplicationConfig, C] = genericReader
})

from grafter.

No difference really... Maybe the time difference just come from producing the AST from inside the macro, not compiling it afterwards.

from grafter.

The main thing that's different between the two macros is that the Reader macro has to figure out the config class name from the type parameter of the annotation. I'm trying to see if that takes an unusually long time. It does contain a couple of loops, but they were intended to break early.

from grafter.

@etorreborre If I understood our last discussion in this topic, this is not an issue anymore, right?

from grafter.

Right this can be closed now.

from grafter.