This would be done in the fast-turn-based system that the game already uses.
I chose to put both [w] and [a] so that I could use them like parentheses, so that the last example on my post would make sense. I suppose we could get rid of that, but I would like to allow it as an alternate control scheme.
Second, six directions is good. And, yes, I'm talking about the vector pointing away from the handle.
I don't think that we need to worry about wrist control for weapons if we don't have rotation around the axis along the forearm. Let me explain: grab a rod. Hold it in front of you, with one hand, pointing straight up and down. This is the "foreward" position. Rotate your wrist to the left or right--not much difference, huh? Now rotate it so that the knuckles of your hand point downward. Hey, look, you're in a position to stab! Rotate your hand so that the knuckles point upward. Your rod should point back along your arm.
Now hold your rod in your first position again, and rotate your forearm. The tip should swing through about a 180 degree arc, with the final position horizontal. That's much more useful than the other rotations we were doing, besides the two vertically inclined ones. If you get your arm into it, you can almost go 360 degrees, which ends up being very, very useful. But it's also very, very stance changing.
Two handed weapons would have an "effective" wrist, caused by twisting the arms around--but if we want arbitrarily accurate positions, it'll probably become a mess.
For now, let's leave the buckler control scheme as an isolated example of what we would like to see in wrist control--we should get an OK comprehensive system rather than getting hung up on rod-like weapons.
Going back to your previous post, what we really need are length measurements for tiles, items, and anatomy. With those, the game should be able to know where everything is in terms of tiles, right?