So I just finished wiring the prototype region editor system wired into the terrain generation code. (As predicted, it’s much too slow to work in real-time for a full region, but is actually mostly okay for the smaller sub-regions within a region. There are lots of improvements that could be made to the speed, too)
But the very first time I modified a bit of terrain using it (pictured at left), I realised that this approach simply doesn’t work — because the user can only modify polygonal boundaries, it ends up generating polygonal-looking terrain.
I should have realised that this was going to happen; it’s almost exactly what happened when I first implemented terrain editing using this sort of system, about three years ago.
There are things I could do to improve it, but they’d just slow the process down even more. Need to step back and think this over a little. Very dispirited at the moment, though; feels like it’s throwing this whole approach to solving the terrain editing interface into question.
Maybe I should throw this whole thing into the “figure this out for version 3” basket — just let the user set a terrain type on each region in version 2, and go with that. Stop wasting time on this intricate terrain editing. Kind of scary to go back in the archives and see that I was working on exactly this same “build the world skeleton” task three years ago, and was running into similar issues even then. (I’ve gotten huge amounts of other things accomplished during that time, of course! But this one has remained a surprisingly large challenge)