Explain geometric constraints solver for CAD to a newbie

[cave_cat]

I have seen the following said on a forum:

suppose I've got a triangle-shaped part, and stick the flat side of it on a square part. What's the distance from the tip of triangle to the opposite flat side of the square? OpenSCAD knows, in some sense, because it drew it. But inside my OpenSCAD script, I don't, and I have to calculate the geometry in my code.

This makes good sense to me. To me, as a programmer, this sounds like he is asking for a better API to look up the coordinate at runtime; perhaps that would force the engine to evaluate expressions in a particular order, but then that's what functional programming is all about. But, in reply people talked about "constraints solver" for this purpose, e.g.:

This is usually done by a constraint solver. The most popular one in the industry is D-Cubed 3D DCM by Siemens.

Can somebody give a simple, toy example based explanation of what this "constraint solver" is about and how it is better than using a better object oriented API to access info about rendered 3d components?

 

[jack action]

Can somebody give a simple, toy example based explanation of what this "constraint solver" is about

What[/PLAIN] is a constraint solver

how it is better than using a better object oriented API to access info about rendered 3d components?

http://www.plm.automation.siemens.com/en_us/products/open/d-cubed/animations/3ddcm/video_3d_spline_constraints.shtml
http://www.plm.automation.siemens.com/en_us/products/open/d-cubed/animations/3ddcm/video_3d_support_cones_tori.shtml

But if it is only for the simple problem presented in the OP, it is like using a 1000 hp engine in a scooter.

 

[cave_cat]

jack action, thank you for the examples. I guess I still don't fully see the "constraint" aspect in them -- these look to me more like propagation of variable change in a functional program. But, for one thing, maybe that's just what they call it in the CAD space, i.e. "constraints" instead of "functional"; and for another, maybe some of this stuff really does go beyond what is normally doable with functional programming and I am simply too clueless in geometry to appreciate that.

 

[jack action]

There is a Wikipedia page about this type of constraint.

The «constraint» aspect is based on the theory of the kinematic pairs in the study of machines:

https://upload.wikimedia.org/wikipedia/en/thumb/3/31/06112014035310LowerKinematicPairs.jpg/512px-06112014035310LowerKinematicPairs.jpg​

 

[timthereaper]

My master's thesis was on CAD software design, so I understand your frustration with CAD APIs in general. The "constraint solver" is simply a numerical method that solves a series of constraints in a CAD assembly model. For example, the human elbow is constrained to a hinge-type motion but only for a certain range of angles. If you were to model the bones in the arm, you would have to constrain the motion of the humerus and the radius/ulna to rotate about the elbow in one axis and only in a certain range. To do that, they create a system of equations and solve it to get the positions of the bones. This helps when trying to flesh out a design for a machine to know what the range of motions are of the parts in the assembly.

The constraint solver is basically a class that uses numerical methods, so you get into the classic problem of properly integrating OOP into scientific computing. I haven't used OpenSCAD, so I don't know about how the API is, but I know that for other CAD systems such as Siemens NX, Dassault Systemes CATIA, and PTC Creo, they're horrible to work with. Those were designed a long time ago and they are often unintuitive and hard to work with.