Stackoverflow

[paulzwu]

With points (1.01, 101.01),(1.01, 101.5),(1.5, 101.5),(1.5, 101.01), it gets stackoverflow. Please check it.

[josdejong]

I have similar issues: the algorithm sometimes becomes unstable and gets stackoverflows.

[josdejong]

FYI: I'm using a different ConcaveHull implementation which works fine for my use case:

https://github.com/atolcd/pentaho-gis-plugins/tree/master/concave-hull

[Pablo1990]

Same issue here. Will try @josdejong solution, which can be also integrated with your pom by adding:

<dependency>
       <groupId>concave-hull</groupId>
       <artifactId>core</artifactId>
	<version>20170808</version>
</dependency>

<repository>
	<id>magsilva</id>
	<url>dav:http://magsilva.pro.br/repo/maven2</url>
</repository>

[rksh]

Since this is a self-calling method (inside a while loop, no less) adding bounds-tests around
https://github.com/Merowech/java-concave-hull/blob/master/ConcaveHull.java#L250
and
https://github.com/Merowech/java-concave-hull/blob/master/ConcaveHull.java#L276

could prevent the overflow, but doesn't solve the fact that the algorithm isn't converging on a solution in iteration at these cases. Investigating line-segment intersection and in-polygon tests.

[rksh]

#3 fixes the test case by re-examining how this implementation breaks ties, and (I think) will prevent stack overflow on unsuccessful exhaustion. Feedback very welcome.