Scala vs Groovy for gradle plugin implementation

[maiflai]

@Kwestor suggested that it would be easier to use Scala to implement the gradle scoverage plugin, as it directly uses some of the classes provided by the scalac-scoverage-plugin (which is itself written in Scala).

I considered using Scala when writing the first version of gradle-scct, but I decided against it because I think that gradle loads all plugins into the 'build' classpath, and this may lead to a Scala version clash between two independent plugins.

I think that the scoverage Maven plugin is written in Java, but if it was written in Scala then I would likewise expect version clash.

I think SBT deals with this by saying that a version of SBT requires a single fixed Scala version, and SBT plugins must be released against a version of SBT; i.e. a new major version of SBT requires all plugins to re-publish, and all plugins must support SBT's current Scala version.

Gradle is more flexible about these things, and this leads to low maintenance overhead for plugin writers. A single release of this plugin supported practically all versions of Scala and Scoverage (until the 1.0 release renamed some of the reporting classes).

There was some discussion in the scalac-scoverage-plugin repository about refactoring the reporting from the sbt-scoverage-plugin, moving it up to the scalac-scoverage-plugin. If this can be done in a Java/Groovy-friendly way (as the current reporting classes are), then I hope the functionality can be neatly re-used from Gradle, Maven and SBT.