Moved to Scalaxy

THIS WORK IS SUPERSEEDED BY Scalaxy/Loops, which is tailored for Scala 2.10.0.

Please DO NOT USE THIS (will only work with Scala 2.10.0-M6 anyway).

What is it ?

This project is a simple yet fully functional prototype to port a subset of ScalaCL / Scalaxy's loop rewrites to Scala 2.10.

The goal is to integrate it into Scala, if possible.

In the meanwhile, it supports foreach comprehensions of special Ranges that are automagically rewritten to while loops:

import scala.inlinable._

for (i <- 100 to_ 0 by -2) {
	println(i)
}

Filters are not supported yet, but nested loops pose no problem (the major problem here is to remember to put that nasty _ suffix to to_ and until_ :-S).

Build / Test

Please do yourself a favor and use paulp/sbt-extras's sbt script (this will help you get sbt 0.12.0-RC3 and Scala 2.10.0-M6 automatically).

Get the sources and run the tests

You can checkout the code with:

git checkout git://github.com/ochafik/optimized-loops-macros.git
cd optimized-loops-macros

And run the tests with:

sbt ~test

(cases for all the variants of start to_/until_ end [by step] are tested, with positive and negative steps)

Build / publish locally

I haven't published scala-inlinable to the Sonatype OSS Repository yet, so if you want to use it you'll have to deploy it locally:

sbt publish-local

Usage example

First, make sure to have published scala-inlinable locally (see above).

Create build.sbt as follows:

scalaVersion := "2.10.0-M6"

libraryDependencies += "com.nativelibs4java" %% "scala-inlinable" % "0.1-SNAPSHOT"

// Uncomment the following option to examine the code after application of macros
// (you'll see while loops instead of Range.foreach calls)
//scalacOptions += "-Xprint:typer"

Put the following in project/build.properties (for paulp/sbt-extras's sbt script):

sbt.version=0.12.0-RC3

Create Test.scala:

import scala.inlinable._

object Test extends App {    
	for (i <- 0 until_ 10) println(i)
	for (i <- 0 to_ 10 by 2) println(i)
}

Run with:

sbt clean run

You'll see the following output that confirms your loops were optimized:

[info] /Users/ochafik/src/macrotest/Test.scala:4: Optimized this loop
[info]   for (i <- 0 until_ 10) println(i)
[info]          ^
[info] /Users/ochafik/src/macrotest/Test.scala:5: Optimized this loop
[info]   for (i <- 0 to_ 10 by 2) println(i)
[info]          ^

What's next ?

Let's see what the feedback is first, in particular from Scala Compiler gurus.

Ideally, this should be merged into Scala (scala.inlinable.InlineableRange should just blend seamlessly into Range).

Also, for ideas of future developments, checkout which optimizations ScalaCL supports, for instance.

Where can this project be discussed ?

• From this discussion on scala-internals
• On Twitter: @ochafik
• Here: fork, modify, submit pull-requests!

Additional notes

Haven't tried to figure things out yet, but the library should have a "self-erasing" behaviour: it's needed at compile time because it bundles its macros, but these macros should erase all traces of the library's types so it won't be needed in the classpath at runtime (to be confirmed, feedback is welcome).