Curve curve intersection #195
Conversation
The LineIntersections struct (and function) may go away after this is implemented, but I’m holding onto it for now.
…rminate properly.
There was a problem hiding this comment.
It's not surprising this doesn't work well - when an intersection is extremely near a subdivision point, then due to floating point roundoff it's possible it can end up in both or neither of the subdivisions.
This is a legitimately tricky problem to get right, the robustness even more so than efficiency. It might be worth taking a look at the lyon code - and also might be good to know that we have permission to adapt that code as long as attribution is given (they adapted some of our flattening code).
| @@ -825,6 +875,71 @@ impl PathSeg { | |||
| .sum() | |||
| } | |||
|
|
|||
| /// Compute intersections between two `PathSeg`s | |||
There was a problem hiding this comment.
I think if we're going to allocate (for the box dyn), it's easier to just return a Vec here.
A non-allocating iterator is also possible, but makes the case analysis harder.
As discussed in #194. Doesn't quite work.