USER FORUM

(you are viewing a thread; or go back to list of threads)

Ellipse? (by Andrew McIntyre)
Apologies if this has been asked before. Is it possible to draw a 2D ellipse in SolveSpace?
Sun Jun 26 2016, 14:43:05
(no subject) (by Jonathan Westhues)
There's no ellipse entity. You can approximate one closely with a spline, of course. (Maybe there should be a generalized conic section entity, like for the optics people or something?)
Sun Jun 26 2016, 14:54:33
(no subject) (by whitequark)
@Jonathan, maybe! Just today I was considering a need to do some pipe miter joints, which that ought to help with...
Mon Jun 27 2016, 07:03:09
(no subject) (by Jonathan Westhues)
What kind of miter joints? You can't sketch an ellipse, but you do end up with one (exactly) when a Boolean cuts a cylinder with an angled plane.
Mon Jun 27 2016, 22:51:59
(no subject) (by whitequark)
A T-junction. Currently you need two cuts, to make a hole in the longer section and a cutout in the shorter, and a fair amount of construction. It is especially painful because of how slow and fragile around lathe groups the geometric kernel is.
Tue Jun 28 2016, 03:46:11
(no subject) (by EvilSpirit)
Ellipse isn't a problem if very want to have it. But there is no point in general conics. Istead, we have to make entity with arbitrary formulas: pos = ExprVector(x(t), y(t), z(t)). We need to approx it with Beziers. Then we get ability to build all kinds of parametric features: parabolic anthenna, aspherical lens and so on.
Tue Jun 28 2016, 04:59:17
Seconded (by Samuel)
I wanted to use solvespace for a scientific project, but the lack of ellipses is a show stopper. I would be delighted if exact ellipses (and possibly other conics) were in the cards for future versions.
Fri Dec 2 2016, 08:44:53
(no subject) (by Paul)
I once made a chess set out of CSG conics. They are very useful. I had a primitive where you specify 3 points, fit a conic through them and revolve it. It was in a ray tracer that supported arbitrary CSG with conics and other things. I still have the code, but perhaps not the chess game. My point is conics can be really useful - all of them, not just ellipses. Splines are probably better though for everything except when an actual conic is required.
Fri Dec 2 2016, 19:18:25
(no subject) (by whitequark)
@Samuel, what sort of project is that?
Sun Dec 4 2016, 14:21:33
(no subject) (by EvilSpirit)
Actually if we add w to free parameters for splines, we get all kinds of rational curves including conics. But actually I don't know local way of constraining w for getting circular, elliptic and other nurbs layout
Tue Dec 6 2016, 09:28:25
(no subject) (by EvilSpirit)
*logical way
Tue Dec 6 2016, 09:29:05
(no subject) (by Paul)
I believe if you take a 1/4 circle represented as a NURBS you can simply stretch it to a 1/4 ellipse by stretching the control points farther apart on one axis. The weights do not need to change. Affine transformations can be applied to NURBS by direct application to the control point locations.
Tue Dec 6 2016, 14:26:28
(no subject) (by EvilSpirit)
Actually there is no problem to make two new entities: ellipse and elliptic arc with correspondent constraints
Sat Dec 10 2016, 07:01:53
(no subject) (by Nithin)
Having seperate ellipse entities is advantageous. Derivatives such as its center point or focii may be of use later in modelling. If we do it using arbitrary formulas: pos = ExprVector(x(t), y(t), z(t)). We might need further functions such as medial axis transform to obtain these derivatives.
Wed Dec 21 2016, 07:18:49
Ellipse? (by P Dii)
I used a spline, though I am not sure it is accurate.
However, I do think an ellipse is actually more important than a circle as with 2 diameters I can create a circle.
Would it be easier to be able to make trace lines "real" lines (the same way as construction lines) then we could contsruct curves using the trace engine.
Sat Jan 21 2017, 20:50:56
(no subject) (by whitequark)
> Would it be easier to be able to make trace lines "real" lines (the same way as construction lines) then we could contsruct curves using the trace engine.

No. Trace lines are piecewise linear, entities are exact.
Sun Jan 22 2017, 04:59:23
Post a reply to this comment: