|
From: | John Swensen |
Subject: | Re: GSOC 2016 Enquiry (boolean operations on polygons) |
Date: | Sun, 24 Apr 2016 22:09:14 -0700 |
Based on my past experience with ClipperLib, it has some issues: 1) It only does integer math. Some of this can be worked around by scaling, using ClipperLib, then scaling the back. The problem with this is that to ensure you are scaling properly, you have to really scan the entire set of points first. This may end up being prohibitive and still may produce incorrect results. 2) For drawing polygons with holes in the past, I have use the poly2tri library (https://github.com/greenm01/poly2tri) with both Boost, gpc, and ClipperLib in the past. It has a non-standard license, but seems like a pretty permissive license. I’m not sure what actually counts as GPL-compatible. I guess I was under the impression we had decided to use Boost, rather than ClipperLib for a variety of reasons (it is still maintained, it handles self-intersections better, it is used by a lot of other GIS/mapping professionals, etc,). I think that implementing this functionality using a new library and writing a suite of tests for regressions, as well as transitioning any other slower functionality of the existing polygon functions, could easily take the full three months. If we want to tack on a few other things, then I guess that could be reasonable. I was going to have a kick-off Google Hangout meeting with him tomorrow to start discussing the timeline and goals, so I guess we should make a decision about the approach we want him to take before we meet. John S. |
[Prev in Thread] | Current Thread | [Next in Thread] |