Coverage of match branches is lost

[sirthias]

I'm using sbt-scoverage 1.0.2 and do not get any fine-grained coverage data on the branches of a @switch match expression. If the match expression is evaluated then all branches are marked green, rather than only the actually used one.

Example:

[sksamuel]

I think what's happening here is its generating two levels of "coverage statement". One is that the match was entered (the highlighting you see), and another for each case. So the fix is to make the latter highlighting override the former. Thanks for bug report.

[sirthias]

Ok, I see. Ideally it'd be possible to even highlight the evaluated matching and guard expressions.
E.g. in this snippet, assuming Extractor1 matches "foo" and Extractor2 doesn't match "foo",

"foo" match {
    case Extractor1(x) if guard1(x) => branch1()
    case Extractor2(x) if guard2(x) => branch2()
    case "foo" => branch3()
    case Extractor3(x) => branch4()
}

I'd expect these expressions to be green:

• the two "foo"
• Extractor1(x)
• guard1(x)
• Extractor2(x)
• branch3()

and these to be red:

• branch1()
• guard2(x)
• branch2()
• Extractor3(x)
• branch4()

[sksamuel]

BTW, It wasn't possible (as far as I could see) to instrument the pattern, this seemed to break the compiler. The guards can be instrumented individually, as can the body blocks.

So in your example, we can do

Green:

the two "foo"
guard1(x)
branch3()

Red:

branch1()
guard2(x)
branch2()
branch4()

[sksamuel]

Can you try release 1.0.4

[sirthias]

Thanks, Stephen, your fix is great! (Sorry for not reacting earlier!)
Things actually highlight nicely now. The match expressions not being highlighted perfectly is not that much of an issue.
Thanks again!

(PS: There is still a problem with not counting pattern matches as "branches" in the report overview, but I'm gonna create a separate ticket for that.)