Constrained parametric design seems like an awesome time saver.. but it's still brittle to the point of being unusable.
Here is a simple example of where it's broken.
1) create user defined parameters, width=64mm, depth=64mm, thickness=0.2mm, and bending_radius=1.5*thickness\
2) create a centered rectangle of dimensions width x depth
3) create a 2-point rectangle from the top-left corner to the top-right that extends upwards by PI*bending_radius/2
4) change the width to 50mm
If this was part of a complex design, it'd break badly (the parameters might as well not be parameters)
Lets say for fun the user wants to find a work-around. They manually put in constraints, and try again, but to no avail.
They manually draw lines extending upward by PI*bending_radius/2 on either side and connect them with another line.
Looks good.
Now they change width to 50mm
The only thing that I've found that actually behaves reasonably is to put another centered-rectangle at the midpoint of the top line in the first, set its dimensions to width x 2* PI*bending_radius/2, but that is unnatural and leaves you with an extra line
.. and if you delete that line and change the width to 50mm
zooming in shows there are still two parallel lines
but they're now separated by some random very tiny value that's definitely not PI*bending_radius/2
This is just one example of where parametrics break, but I've never been able to make parametrics work for anything nontrivial (read: more than a few geometric shapes).
This desperately needs to be fixed because in its current form, it's worse than useless because being there and listed as a feature, people might actually try to use them, and in a complex design, they'll at best break in a visible way (merely frustrating the user and wasting their time), and, at worst, they'll break in an invisible way and the user will spend money, time, and resources making something only to find that the geometry.is wrong (and wasting more time just debugging it).