The future of scala metaprogramming

# The Future of Scala Metaprogramming

http://goo.gl/F29sd9

---

# Talks

* ⇒ Metaprogramming 2.0 - Eugene Burmako https://youtu.be/wii5UPtu1_g (Scala Days 2016 NYC) http://goo.gl/xUUxOE (Slides Scala Days 2016 Berlin)
* ⇒ Dotty Linker: <code>...</code> - Dmitry Patrashko https://youtu.be/h8KBLF0AgUc (Scala Days 2016 NYC) http://goo.gl/XrGGtH (Slides Scala Days 2016 Berlin)

* Scalan: <code>...</code> - Alexander Slesarenko
* Data centric Metaprogramming - Vlad Ureche - https://youtu.be/83VfxITTNPE (Spark Summit Europe 2015)

???

Scalan: DSL development/compilation framework

Vlad Ureche: using meta-programming to speed up applications

---

# Macros

```
type Macro = Code ⇒ Code
```

---

# Macros

```
type Macro = Code ⇒ Code
```

* Treating programs as data
* Transform programs
* Compile-time reflection
* Extending the language/compiler

---

# Macro support

2.10: Macro's introduced as experimental feature

---

# Macro support

2.10 - 2.11: Whitebox macros

2.11: Blackbox macros

???

Original plan was to stabalize blackbox macros

---

# Scala.meta

![New and improved](newimproved.jpg)

---

# Macro support

2.10 - 2.12: Whitebox macros

2.11 - 2.12: Blackbox macros

2.10 - 2.12: scala.meta

Dotty: scala.meta

Scala Native: scala.meta

Scala.JS: scala.meta

???

---

# Scala.meta

* Easier and less boilerplate

---

# Scala.meta

* Easier and less boilerplate
* Compiler implementation independent    dotty, scala native, scala.js

---

# Scala.meta

* Easier and less boilerplate
* Compiler implementation independent    dotty, scala native, scala.js
* IDE support

---

# Scala.meta

* Easier and less boilerplate
* Compiler implementation independent    dotty, scala native, scala.js
* IDE support
* Improved parsing (dialects, formatting)

---

# Scala.meta roadmap:

* Typechecking, name resolution, implicit interference
* Inline macros
* 2.10, 2.12, scala.js

---

# Example: @main

```
@main object Test {
   println("hello world")
}
```

---

# Example: @main

```
@main object Test {
   println("hello world")
}
```

↓

```
object Test {
  def main(args: Array[String]): Unit = {
    println("hello world")
  }
}
```

---

# Example: @main scala.reflect

```
class main extends scala.annotation.StaticAnnotation {
  def macroTransform(annottees: Any*): Any =
    macro Macros.impl
  }

import scala.reflect.macros.whitebox.Context
object Macros {
  def impl(c: Context)(annottees: c.Tree*): c.Tree = {
    import c.universe.
    val q"object $name { ..$stats }" = annottees.head
    val main = q"""
    def main(args: Array[String]): Unit = { ..$stats }
    """
    q"object $name { $main }"
  }
}
```

???

Bad: boilerplate: same amount
Good: quasiquotes

---

# Example: @main scala.meta

```
import scala.meta._
class main extends scala.annotation.StaticAnnotation {
  inline def apply(defn: Any) = meta {
    val q"object $name { ..$stats }" = defn
    val main = q"""
      def main(args: Array[String]): Unit = { ..$stats }
    """
    q"object $name { $main }"
  }
}
```

---

# Example: @main scala.meta

```
@main object Test {
   println("hello world")
}
```

---

# Example: @main scala.meta

```
object Test {
  def main(args: Array[String]): Unit = {
    println("hello world")
  }
}
```

---

# Other macro uses

* Scalafmt
* Automatic serialization ([Play JSON](https://github.com/playframework/playframework/blob/afd06b38241e158229393ab68724a2cbd82d444d/framework/src/play-json/src/main/scala/play/api/libs/json/JsMacroImpl.scala), [Pickling](https://github.com/scala/pickling/blob/4f5dd03ae103c4e6d42eae1c14b3180a1813bae4/core/src/main/scala/scala/pickling/generator/PicklingMacros.scala))
* Debugging (e.g. [print all variables in scope](https://github.com/adamw/scala-macro-debug))
* Performance improvements

---

# Performance improvements

```
if(logger.isErrorEnabled) {
  logger.error(expensiveArgument());
}
```

⇔

```
logger.error(expensiveArguments());
```

---

# Performance improvements

```
val arr = (1 to 10000000).toArray
arr.reduce(_ + _).toDouble / arr.length
```

⇔

```
5000000.5
```

???

It will take scala about a second to determine that the above is indeed that number

---

# Dotty TASTY

* Typed AST
* Binary distribution format for dotty

???

Distribution format of Dotty scala binaries

---

# Dotty linker

Whole-program optimization

* Dead code elimination *for your specific usage*
* Specialization
* Inlining
* Virtual dispatch removal
* Performance rewrites

???

Specialization normally not viable: 10^n combinations n= program args

---

# Rewrite rules

```
import dotty.linker._

@rewrites
object LoggerOptimization {
  def error(x: Seq[Int]) =
    Rewrite(from = logger.error(s),
            to   = if(logger.isErrorEnabled) logger.error(s) else ())
```

???

Readability or performance: why not both?

---

# Rewrite rules

```
import dotty.linker._

@rewrites
object LoggerOptimization {
  def error(x: Seq[Int]) =
    Rewrite(from = logger.error(s),
            to   = if(logger.isErrorEnabled) logger.error(s) else ())
```

Correctness: automatic scalacheck tests (planned)

---

# Summary

##### Scala.meta

* simplifies meta-programming in Scala
* Only macro system of dotty, scala native, scala JS
* High-fidelity parsing
* No feature parity yet

---

# Summary

##### Scala.meta

* simplifies meta-programming in Scala
* Only macro system of dotty, scala native, scala JS
* High-fidelity parsing
* No feature parity yet

##### Scala linker

* Whole-program optimization
* Rewrite rules (performance oriented)

---

# Summary

##### Scala.meta

* simplifies meta-programming in Scala
* Only macro system of dotty, scala native, scala JS
* High-fidelity parsing
* No feature parity yet

##### Scala linker

* Whole-program optimization
* Rewrite rules (performance oriented)

http://goo.gl/F29sd9