I've been trying to stick with Eclipse now for a while. I switched from IntelliJ 8 to Eclipse because IntelliJ stopped working for Tapestry (I still don't know why) and because my fingers were getting tied in knots switching between IntelliJ and Eclipse. I have to switch back and forth because my Tapestry training uses Eclipse and I got a lot of negative feedback the times I tried to get people to use IntelliJ as part of the training.

In any case, I've been gritting my teeth against the travesty that is Eclipse; the inconsistent behaviors, the needless complexity, the lack of sense in much of the UI. I often wonder what IDE the Eclipse developers use, because it couldn't possibly be Eclipse itself, or they would have fixed some of the more brain-damaged stuff years ago.

One thing that is currently killing me is that Eclipse has no concept that some source folders contain production code and some contain test code. Because of that, when I collect coverage, EMMA instruments the test code as well as the production code. That not only throws off coverage figures, but breaks some of my tests (advanced ones about bytecode manipulation, that are sensitive to when EMMA adds new fields or methods to existing classes).

Fortunately, EMMA has an option to restrict its instrumentation (though it a global preference, and not configurable for individual projects) ... but shouldn't Eclipse understand this distinction natively? IntelliJ does, and it helps prevent a lot of problems, not just with coding, but with testing as well.

I keep hoping that there's a better, faster, simpler solution out there ... something that is elegant and precise. Eclipse is dumb as a sackful of hammers, and IntelliJ is almost fractal in its complexity, and ugly to boot.

I haven't found my perfect IDE yet. Maybe it's NetBeans?

From http://tapestryjava.blogspot.com/2010/12/is-it-time-to-switch-back-to-intellij.html