It looks very good, but I need a constrained Delaunay triangulation solution. I also use libTess for its robust boundary contour operations for clipping and booleans.
-
-
Replying to @Vector_GL @SasMaster1980 and
For what it’s worth, I like the libtess approach of monotone polygon decomposition a lot better than ear clipping. I just have an engineer’s bias toward simplicity, which is why I like trapezoidation :)
1 reply 0 retweets 0 likes -
Replying to @pcwalton @SasMaster1980 and
libTess is academically impeccable - but it produces a lot of tall very skinny triangles due to the horizontal sweepline.
2 replies 0 retweets 0 likes -
Replying to @Vector_GL @SasMaster1980 and
Ditto with trapezoidation. In fact I believe trapezoidation plus monotone polygon decomposition is *equivalent* to the libtess approach.
1 reply 0 retweets 0 likes -
Replying to @pcwalton @Vector_GL and
In other words: Lorenzetto trapezoidation is half of libtess. The problem I had is that monotone polygon decomposition is really hard to make work, maybe impossible, while preserving curves. So I just threw that part out. End result: A simple approach.
3 replies 0 retweets 0 likes -
Replying to @pcwalton @SasMaster1980 and
I suppose Pathfinder has to deal with a lot of skinny horizontal triangles instead ;-) At one stage I considered a post triangulation merging algorithm.
1 reply 0 retweets 0 likes -
Replying to @Vector_GL @SasMaster1980 and
Yep, it does have a lot of skinny triangles. But I’m willing to accept that drawback because it’s so simple. I want to ship and have the code be maintainable :)
2 replies 0 retweets 0 likes -
Replying to @pcwalton @SasMaster1980 and
Also, I’ve yet to find a triangulation solution that has acceptable latency. I can run my fat scanline algorithm is a few ms. Triangulation can easily take 10x longer. All those linked lists are so cache hostile.
2 replies 0 retweets 0 likes -
Replying to @Vector_GL @SasMaster1980 and
Yep, same results here! I actually removed the red-black tree I was originally using for the active edge list. The resulting O(n^2 log n) algorithm is actually faster in practice than the O(n log n) one.
1 reply 0 retweets 0 likes -
Replying to @pcwalton @Vector_GL and
You guys over complicating everything ;) for text rendering I tried
@EricLengyel and I won't spend my time from now on reinventing the wheel. For shapes loop blinn + kojojima mix is simple, fast and robust for absolutely most of real world use cases.2 replies 0 retweets 0 likes
Well, Pathfinder’s approach is simpler than that of Slug. See the shader: https://github.com/pcwalton/pathfinder/blob/master/shaders/gles2/stencil-aaa.fs.glsl …
-
-
Replying to @pcwalton @Vector_GL and
Yeah, but why Rust? ;) I actually have been following your work, but I have no plans to adopt or port from Rust at the moment. Maybe some good people with lots of free time will eventually do it.
0 replies 0 retweets 0 likesThanks. Twitter will use this to make your timeline better. UndoUndo
-
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.