Improvingworks Infinitest

from Daniel Yukio Yokomizo:

Is it in the plans or there's a way to integrate Infinitest with code coverage tools? Currently I'm using EclEmma to assess coverage inside Eclipse and thought it would be a good idea to have this information as Infinitest runs the tests, for example if I change some code and the tests succeed but coverage is worse it would be nice to have a warning on the new, uncovered, line.

from Ben Rady:
Oh, we've got plans...it's the implementation that eludes us :-)

The one tricky bit is that because Infinitest is selectively running tests, it can't really tell you what code is uncovered. It can definitively tell you that a particular change covered a specific set of lines, or statements, but in terms of your overall coverage for the entire test suite, we can't really tell you that a line is "uncovered". Is this actually a problem? I have no idea.

from Daniel Yukio Yokomizo:
Can't Infinitest establish a baseline at some point (e.g. when a explicit build is invoked or implicitly like clean+build automatically )? Just a thought.

from Ben Rady:
Also, how you represent this information is very important. I think the red/green line highlighting that you see in most coverage tools is awful. Makes my eyes bleed every time I look at the screen. One thing I've thought about is a subtle hint in the editor that lets you know a particular statement is covered. Turning the semicolon green, perhaps.

from Daniel Yukio Yokomizo:
I was thinking about a warning on lines that are new and uncovered. If a line is new or was changed and it wasn't covered by the test a warning is placed on the line. Other lines should be ignored. If I later write a new TestCase that covers the line marked with the uncovered warning, it should be cleared. AFAICS this will never show false positives (i.e. lines that are covered but show as uncovered) but it may cause false negatives (i.e. uncovered lines without warnings). As usual the user can always invoke a full build (forcing all tests to run), so the complete coverage would be assessed. IMHO Infinitest works best when I can immediately see if I introduced a bug in the program (essentially I care about what I just did not about the rest of the program), so a coverage feature that shows only if my last change is uncovered is enough (for me).