Addictive Software

var me = Random.shuffle(List("Developer", "Cook", "Geek", "Musiclover", "Engineer"))

Code coverage in scala

Most code coverage tools check lines of codes and branches (which part of the if/then/else is covered)

For scala this is not a very useful metric, i'll let the following example speak for itself

Players.filter( == true).sortBy(- _.score).take(5)
   .foreach(p => println(s"${p.lastname} ${p.score}"))

in this case it is all on one (or two) line(s) and any branching (f.i. in the .filter()) is hidden inside the collection logic

Therefor the best metric to use for scala is statement coverage, meaning the coverage tool will look at each statement, f.i. == true and - _.score etc.

scoverage is a githbub project that provides statement and branch coverage for scala code, it comes with sbt, maven and gradle plugin and its reports can be used in sonar or coveralls

an example using sbt:

in my project/build.sbt I added

resolvers += Classpaths.sbtPluginReleases

addSbtPlugin("org.scoverage" %% "sbt-scoverage" % "")

and then in build.sbt


now you can run

sbt scoverage:test

data and reports are available in target/scala_2.10/scoverage* directories

At my job we use Team City as our Continuous Integration tool and it only supports a couple of code coverage tools, so it means we can't fail a build when the coverage falls beneath a certain number.

What you could do is add the html report as artifact so at least people can look at the build and click a link to see the report
to do this add the following to your "publish artifacts" section of your build:


To conclude code coverage is not an end-all-solution for code quality, but it surely helps.
And try to resist the urge to add unit-tests just to get the coverage up, think about the functionality that you are testing and what's missing there.

comments powered by Disqus