It all just seems kind of pointless to me.
Motion is quantized on the level of tiles in DF spatially, and turns temporally. No reason to make it look like it's any more precise than that, I don't think.
That's what I said to my buddy originally, actually (well, I didn't use those exact words) but he was excited about the idea and pseudo-convinced me and I said I'd put it up.
If you're running at the default 100 FPS and 20 GFPS, that is 5 frames per graphics update.
Assuming that said dwarf moves every frame, which is possible with an agile dwarf, you'll have him move here before the graphic updates.
. . D . O
.XX X X X
.X D - Dwarf
.X O - Destination
.X X - Wall
Would the OP please care to explain how tweening would work in this instance?
It's possible to move *EVERY* frame with an agile dwarf? Uh, not unless it has SPEED:0...
But even so, the point was that long-distance updates can screw up the "movement" if there are turns involved in the path.
*shrug* I dunno, I don't really care. So for one 40th of a second (assuming 20GFPS, 1 interstitial frame) the D is clipping about 1/4 of the way into a wall. It still makes it easier to follow him with your eyes than if he teleports 5 squares at a time.
The obvious way around it would be to keep track of previous locations since the last G_frame. Still wouldn't affect the DFMA movie renderer for the same reason the original suggestion didn't. But honestly, I wouldn't care whether the "correct path" was followed or not.