Scala 2.7.4-rc1 released Thursday, Apr 16 2009 

Scala 2.7.4-rc1 has been released. It incorporates a few important fixes related to serialization of anonymous functions and generation of Java generics, but the big change is the improved Scala plugin Scala IDE for Eclipse. This means that my backport is now obsolete.

The announcement contains all the juicy information.

MercurialEclipse 1.3.1019 Wednesday, Feb 18 2009 

With the talk about DVCSs as active as ever in the Eclipse world, I thought I’d mention the release of MercurialEclipse 1.3.1019. It still lacks quite a bit of polish, but it already supports a decent amount of features.

Even though Mercurial doesn’t have as much momentum as Git, it has been adopted by a respectable amount of projects (including Mozilla, Netbeans and OpenJDK).

Scala plugin for Eclipse snapshot r17139 Wednesday, Feb 18 2009 

The snapshot of the Scala plugin for Eclipse that I posted a few days ago had a couple of annoying bugs. Here’s a new one build (note that it’s an Eclipse update site, so no point in trying to access it with a browser) from revision 17139 of the trunk of the plugin with the compiler and standard libraries from the 2.7.x branch.

Miles has been busy once again and fixes included since the last build are 1717, 1233, 1636, 1662, 1716, 1627, 1534, 1410, 1516, 1483 and 1253.

Update: New build published that includes fix for ticket 1741. Uninstallation followed by installation may be required if p2 starts giving cryptic errors (it seems like it doesn’t handle well the case where the update site is replaced). Sorry about that.

Update 2: With the release of Scala 2.7.4-rc1, this backport is obsolete.

Scala plugin snapshot Saturday, Feb 14 2009 

Miles Sabin has recently committed the Equinox Aspects-based approach to JDT integration for the Scala plugin to trunk. This approach is similar to the one used by the AJDT as described by Andrew Eisenberg.

Soon after that commit, Miles embarked on a bug-killing mission. Nightlies are available from the usual place, but they rely on the trunk of the compiler and standard library. With APIs still in flux, it would be nice to be able to use the improved plugin with the compiler and standard library from the 2.7.x branch (this includes a few important fixes added after the 2.7.3 release).

Since it seemed like no-one else had time to do that in the near future, I did it myself. It’s available here in case it’s useful for anyone else. In the interest of full disclosure, I applied the following patch to the plugin:

--- META-INF/MANIFEST.MF (revision 17110)
+++ META-INF/MANIFEST.MF (working copy)
@@ -34,8 +34,8 @@
- scala.library,
+ scala.library;bundle-version="[2.7.3,2.7.4)",
Index: src/scala/tools/editor/Typers.scala
--- src/scala/tools/editor/Typers.scala (revision 17110)
+++ src/scala/tools/editor/Typers.scala (working copy)
@@ -6,7 +6,6 @@


-import scala.annotation.unchecked.uncheckedStable
import scala.collection.jcl._

import nsc.{util,io}
Index: src/scala/tools/editor/TypersPresentations.scala
--- src/scala/tools/editor/TypersPresentations.scala (revision 17110)
+++ src/scala/tools/editor/TypersPresentations.scala (working copy)
@@ -6,7 +6,6 @@


-import scala.annotation.unchecked.uncheckedStable
import scala.collection.jcl.{LinkedHashMap,LinkedHashSet}
import scala.collection.mutable.ListBuffer

@@ -595,10 +594,13 @@
val str = name.decode
val key = if (sym.isMethod) str.trim+header(sym) else str.trim
val last = str.last
+ import

// TODO: check accessibility.
- if (name.isTypeName == isType && str.toLowerCase.startsWith(leading.toLowerCase) &&
- (str.indexOf('$') == -1) && last != ' ' && !contains(key)) {
+ if (name.isTypeName == isType && (str.indexOf('$') == -1) && last != ' ' &&
+ !contains(key) && (str.toLowerCase.startsWith(leading.toLowerCase) ||
+ SearchPattern.camelCaseMatch(leading, str))) {
val sym0 = if (verify==null) sym else verify(sym)
if (sym0 != compiler.NoSymbol) {
val high = if (pt != null) sym0.tpe <:< pt

It includes my simple camel-case content assist patch, some version restrictions and the removal of the imports for the uncheckedStable annotation since it was moved outside of the scala package in 2.8.0. The compiler and standard library are the latest version from the 2.7.x branch. I’ve done some light testing and the plugin works for me, but the usual disclaimers apply. Be prepared to have to uninstall and reinstall the plugin if things don’t work out.

Update: I noticed that two annoying bugs were introduced on Friday. I would recommend not using the current build. I will update it once they are fixed.

Update 2: See here for a newer build with fixes for the mentioned issues.

Update 3: With the release of Scala 2.7.4-rc1, this backport is obsolete.

Code reuse by copy and paste Friday, Feb 13 2009 

I was investigating what would need to be done to implement camel-case code assist for the Scala plugin when I ran across the following gem in CharOperation (Eclipse JDT):

public static final boolean camelCaseMatch(char[] pattern, int patternStart, int patternEnd, char[] name, int nameStart, int nameEnd, boolean samePartCount) {

	/* !!!!!!!!!! WARNING !!!!!!!!!!
	 * The content of this method has been fully copied to
	 * SearchPattern#camelCaseMatch(String, int, int, String, int, int, boolean).
	 * So, if current method is modified, do NOT forget to copy again its content
	 * to SearchPattern corresponding method!

And in SearchPattern:

public static final boolean camelCaseMatch(String pattern, int patternStart, int patternEnd, String name, int nameStart, int nameEnd, boolean samePartCount) {

	/* !!!!!!!!!! WARNING !!!!!!!!!!
	 * The algorithm of this method has been entirely copied from
	 * CharOperation#camelCaseMatch(char[], int, int, char[], int, int, boolean).
	 * Array lengths have been replaced with call to {@link String#length()} and
	 * array direct access have been replaced with call to {@link String#charAt(int)}.
	 * So, do NOT modify this method directly to fix any bug but modify first the
	 * corresponding CharOperation method and do the copy again to be sure that
	 * these two methods are kept synchronized.

I am sure there is a reason why this was done, but how ugly…

Update: At least JDT made it easy to implement what I wanted.