In the attached example drawing using the latest stable QCAD release 3.32.4:
- Try removing segment A with D2 ... Undo (OO)
- Try removing segment B with D2 ... Undo (OO)
- Try removing segment C with D2 ... Undo (OO)
- Try removing segment E with D2 ... Undo (OO)
This doesn't remove only the segments A, B, C or E, the action removes A+B+C+E.Also:
- Try trimming segment A with AX ... Undo (OO)
- Try trimming segment B with AX ... Undo (OO)
- Try trimming segment E with AX ... Undo (OO)
This doesn't auto trim the intended segments A, B or E, the action keeps A+B+E.- - - - - - - - -
The problem here is one of the overall scale, but keep in mind that the ellipse is about the size of a wide doorway.
On purpose the drawing unit is the Kilometer and the numbers are thus quite small
All things are relative ... Know that the moon's orbit major radius is only about factor 1.6 larger expressed in AU as drawing unit.
The blue lines are perfectly tangent to the ellipse and have 1 intersection point with the ellipse.
As pointed out in topic 11900, a line crossing the ellipse twice at two points 'really close together' is considered as almost tangent.
This turns out to be falsely true for the two green lines.
Their intersections with the ellipse are 0.000766 units apart, which is more than 75 cm.
Then the order of the 2 returned intersection points matters.
When the second result is about the same as the first within 0.001 units, the second result is removed from the list.
In this case the intersection point between A and E and the intersection point between E and C.
The intersections with the blue lines are closer to the indicated positions than the remaining and the ellipse is trimmed at the blue lines.
Discarding one of two almost equal solutions is seen more in the QCAD code.
When 'really close together' a point halfway between the two results could be a better solution.
But here that will be on the ellipse major axis and not on the ellipse circumference.
The green lines are no tangents to the ellipse to start with.
When comparing sizes then the same fixed tolerance for all drawing units tend to produce false positives.
'Within 0.001 units' is a 25.4 times larger tolerance expressed in inch compared with mm.
The better option here would be to define 'really close together' in the order of the ellipse size.
For example as 5ppm of the sum of the radii. (Analog of tol = (r1+r2)/200000 for touching circles)
But that may fail just the same as with touching circles when the difference between the radii is too large.
Regards,
CVH