Sice SW didn't think to include rack and pinion mates along with gear mates, I've had to roll my own. It is useful to simulate any transformation from rotary motion to linear motion (or vice versa), and all that is required is to give up a little mechanistic fidelity. I create a 'Screw Surface' component which is nothing but a cylindrical extrusion with a diameter much larger than the size of the components to be mated. With a couple of mates you can constrain the 'rack' of the rack and pinion such that a point on the rack moves along the circumference of the 'Screw Surface' and the screw surface is mated tangent to the pinion surface. In reality, the rack moves in a pendulum fashion, but by making the radius of the 'Screw Surface' very large, the motion of the rack is close enough to linear. With a little math you can calculate the proper ratio for a gear mate which gives you realistic rack and pinion or screw motions.
The problem is that the large bounding box of the 'Screw Surface' component wreaks havoc in the assembly. The proper behavior is that the bounding box for an assembly is governed by the visible comonents in the assembly. If the large component is visible, the bounding box is large. If the component is hidden, the box is small. This behaves as expected, until you put the assembly inside another assembly. Seemingly at random, the subassembly bounding box will be calculated as though the large component is visible. When you autoscale the main assembly, everything becomes tiny. Rebuilding doesn't help, nor does CTRL-Q. If you go back to the subassembly model and autoscale it, it becomes very small as well. CTRL-Q fixes the problem in the subassembly (temporarily) and on return to the main assembly it behaves again, but only for a while.
I have a main assembly with 4 subassemblies, each of which has one of these screw mates and I spend almost as much time fixing this problem as I do accomplishing actual work on the main assembly.
Jim S.