I kind of agree, but this is partially due to accuracy of number used and the fact that coordinates of all points are independent in the solver, so it is hard to detect if the constraints are really redundant or just very close to each other, but conflicting, and unsolvable. Some very smart hybrid (algebraic + numeric) solver could detect most of the common redundancies and eliminate them, point out which ones redundant, or just continue to work without problem. That would be kind of nice, but that can lead people to adding way too many constraints, and not think what they are doing. I know it is currently very normal to deal with redundant or unsatisfied systems all the time, and it is kind of a game with solvespace we pay for it being nice parametric system. I have no idea how other parametric modeling CADs deal with it.

No problem to analyze redundant constraints. Actually, we already used same approach while we adding new constraints with dimension values. If they cause overdefinition of equation system we just mark them as reference automatically after addition. The problem is what we have to avoid using solving by substitution during process of redundant analysis, so this can cause performance issues. But now we have pending improvements about performance and timeouting redundant constraints computations. @whitequark can answer the question about when we can merge this new imorovements.

@whitequark can answer the question about when we can merge this new imorovements.

The answer is: now! :)

Fixed in 5495659.

@Evil-Spirit I've also added this logic to recognize newly added redundant constraints based on rank, what do you think about it? cf2f0e5

Just a bit of a report, I really like this new feature, and it works really well. It reduces number of time I get over constrainted sketches probably by a factor of 3. :)

Thanks.

@whitequark good thing )