sketcher: wrong redundancy solving with symmetry

[uwestoehr]

In recent discussions about the sketcher symmetry handling I mentioned an issue I think it is incorrect. Here it is:

- open the attached sketch:

Sketch-symmetry-issue.FCStd

(9.82 KiB) Downloaded 107 times

The solver tells you there are redundant constraints.

- now delete the symmetry constraint

The redundancy message is now gone but you get 1 degree of freedom. To eliminate it it is sensible to set the symmetry constraint you just deleted.
It is in my opinion a bug that one cannot set here the symmetry. (I am aware that there are workarounds but for the later development I need a symmetry constraint here, not a dimensional one.)

Can anybody reproduce this?

OS: Windows 10
Word size of OS: 64-bit
Word size of FreeCAD: 64-bit
Version: 0.19.17331 (Git)
Build type: Release
Branch: master
Hash: cf18161dc26267c259ac21675cb9637263bf4152
Python version: 3.6.6
Qt version: 5.6.2
Coin version: 4.0.0a
OCC version: 7.3.0
Locale: German/Germany (de_DE)

[openBrain]

Sorry but (IMO) you're wrong and the solver is correct.
The top arc of your example is redundantly constrained because it's center is coincident to the origin.
This geometrically imposes that :

• If its both endpoints are on the same horizontal, the horizontal distance from each endpoint to Y axis is equal
• Oppositely, if its both endpoints are at the same distance from Y axis, they are on the same horizontal (modulo 90° to be strictly precise)

The problem is that a symmetry against Y axis imposes both conditions (same horizontal + same distance to Y axis) => I guess you perfectly understand it's redundant.

I see 2 simple ways of being geometrically correct :

• Either only apply an horizontal constraint between the 2 arc endpoints => IMO this is the best way
• Or keep the symmetry, remove the coincidence between arc center & origin, then only constraint the arc center to be on the X axis with a point-on-line

Hope it helps.

[openBrain]

What baffles me is that solver doesn't complain with the following simple example (arc center is coincident with origin) :

simple_arc.png (2.74 KiB) Viewed 1264 times

At least with my version :
OS: Windows 7
Word size of OS: 64-bit
Word size of FreeCAD: 64-bit
Version: 0.18.16110 (Git)
Build type: Release
Branch: (HEAD detached at upstream/releases/FreeCAD-0-18)
Hash: f7dccfaa909e5b9da26bf50c4a22ccca9bb10c40
Python version: 3.6.6
Qt version: 5.6.2
Coin version: 4.0.0a
OCC version: 7.3.0
Locale: French/France (fr_FR)

[openBrain]

More precisely :

• Sketch an arc with its center coincident to origin => Solver says 3 DoF (OK)
• Add a symmetry between its 2 endpoints against the Y axis => Solver says 1 DoF (NOK)
• Drag the arc with mouse => Solver says 2 DoF (OK !!) but doesn't complain (NOK)

Can someone confirm (preferably with 0.19 version) ?

[bejant]

Add a symmetry between its 2 endpoints against the Y axis => Solver says 1 DoF

Here my Solver says 2 DOF remaining.

20190716a.png (36.55 KiB) Viewed 1251 times

OS: Ubuntu 18.04.2 LTS (ubuntu:GNOME/ubuntu)
Word size of OS: 64-bit
Word size of FreeCAD: 64-bit
Version: 0.19. <== date is 2019/07/08
Build type: Release
Python version: 3.6.8
Qt version: 5.9.5
Coin version: 4.0.0a
OCC version: 7.3.0
Locale: English/UnitedStates (en_US)

[uwestoehr]

I guess you perfectly understand it's redundant.

Thanks for having a look. But I cannot follow your argumentation because when I delete the symmetry constraint I get 1 DOF. Then I can change the horizontal and y-position. An example:

Sketch-symmetry-issue2.FCStd

(12.02 KiB) Downloaded 52 times

But as you write: "symmetry against Y axis imposes both conditions (same horizontal + same distance to Y axis)" and this is exactly what I need.

I see 2 simple ways of being geometrically correct :

• Either only apply an horizontal constraint between the 2 arc endpoints => IMO this is the best way
• Or keep the symmetry, remove the coincidence between arc center & origin, then only constraint the arc center to be on the X axis with a point-on-line

The second one is intuitive and requires additional work -> breaking workflow. The first one is not an option because I need the sypmmetry (sketch will later CCed, arc cut etc.)

What baffles me is that solver doesn't complain with the following simple example (arc center is coincident with origin) :

Same here and that fits perfectly to my understanding of the the symmetry constraint .

[uwestoehr]

Can someone confirm (preferably with 0.19 version) ?

With 0.18 not 0.19 I get correctly only 2 DOF.

OS: Windows 10
Word size of OS: 64-bit
Word size of FreeCAD: 64-bit
Version: 0.19.17331 (Git)
Build type: Release
Branch: master
Hash: cf18161dc26267c259ac21675cb9637263bf4152
Python version: 3.6.6
Qt version: 5.6.2
Coin version: 4.0.0a
OCC version: 7.3.0
Locale: German/Germany (de_DE)

and

OS: Windows 10
Word size of OS: 64-bit
Word size of FreeCAD: 64-bit
Version: 0.18.16117 (Git)
Build type: Release
Branch: releases/FreeCAD-0-18
Hash: dbb4cc6415bac848a294f03b80f65e888d531742
Python version: 3.6.6
Qt version: 5.6.2
Coin version: 4.0.0a
OCC version: 7.3.0
Locale: German/Germany (de_DE)

[bejant]

The solver tells you there are redundant constraints.

- now delete the symmetry constraint

The redundancy message is now gone but you get 1 degree of freedom. To eliminate it it is sensible to set the symmetry constraint you just deleted.

My Solver messages report that the Sketch is Over Constrained (not that there are Redundant Constraints).

Other Constraints should be favored over the Symmetry Constraint so to me it would be more sensible to:
1. Apply the Equality Constraint to the two vertical edges, or less desirably,
2. Apply the Symmetry to upper right and lower left vertices, or to upper left and lower right vertices, and the Y-Axis.

Interestingly, if I apply the Symmetry Constraint before the 4mm Horizontal Distance Constraint, there is no problem (one of the Conflicting Constraints probably got deleted).

(I had to make sure I was dealing with the original file)

[openBrain]

Add a symmetry between its 2 endpoints against the Y axis => Solver says 1 DoF

Here my Solver says 2 DOF remaining.

Thanks for testing with 0.19. I expect some things have move on between 0.18 & 0.19. But still the solver doesn't complain, which is weird.

My Solver messages report that the Sketch is Over Constrained (not that there are Redundant Constraints).

Yep. It depends on whether you use 0.18 or 0.19. I think both are acceptable.

Thanks for having a look. But I cannot follow your argumentation because when I delete the symmetry constraint I get 1 DOF. Then I can change the horizontal and y-position.

No. There is only one DoF because x & y positions are linked by an equation. You can see the (only) remaining DoF as being the arc apex angle, or the arc chord length.

But as you write: "symmetry against Y axis imposes both conditions (same horizontal + same distance to Y axis)" and this is exactly what I need.

The question here isn't what you need. The solver don't care of that.
The question is that constraint set is mathematically (precisely geometrically) correct.
And in this case (an arc with center coincident with origin and with endpoints symmetric against Y axis) there is clearly some redundancy.

The second one is intuitive and requires additional work -> breaking workflow. The first one is not an option because I need the sypmmetry (sketch will later CCed, arc cut etc.)

Unintuitive
With both of them you have the symmetry. Not displayed for sure, but geometrically imposed by constraint set.
Unfortunately their is no "intuitive" constraint set where you can get the symmetry constraint used as is.

Same here and that fits perfectly to my understanding of the the symmetry constraint .

Here again this isn't about how you figure out a symmetry but about what mathematical relation it implies.
If you want to get some idea, sketch an arc and just constraint its 2 endpoints with a symmetry against Y axis(nothing else).
Then try to wiggle everything (the arc, its endpoints, its center) => You absolutely can't have a valid solution where the center is out of the Y axis. Thus any other constraint that will also constraint the center on the Y axis (typically coincidence to origin is one of them) is redundant.

[uwestoehr]

My Solver messages report that the Sketch is Over Constrained (not that there are Redundant Constraints).

Yes, sorry for being incorrect.

1. Apply the Equality Constraint to the two vertical edges, or less desirably,
2. Apply the Symmetry to upper right and lower left vertices, or to upper left and lower right vertices, and the Y-Axis.

1. is no option because I need the symmetry for later actions (sketch will be CCed and the arc will go in the new sketch...)
2. leads here to an over constrained error.