Hello. Le sam. 16 nov. 2019 à 05:40, Matt Juntunen <matt.juntu...@hotmail.com> a écrit : > > Gilles, > > The old commons-math code produces the same results.
:-( > So, I'm not sure if this is a bug or a known limitation. There isn't any > documentation in the old code in regard to the barycenter computation and > there are only two trivial unit tests for it. I had a look, but it's not even a clear to me what trivial cases it is testing. It would help (me, at least) if there were separate methods that indeed spell out some trivial cases (where the expected answer should hopefully be obtained with ultimate accuracy, whatever the implementation). > I suppose if there any no further insights into this, I could just stick with > the existing algorithm, use a large epsilon in the unit tests, and leave any > fixes for future JIRA issues. Yes, but please: * add code comment in the unit tests (as to the unexpected large tolerance required for them to pass), * add Javadoc in the "main" code (signalling that the unit tests point to a potential bug), * file a JIRA report for this issue (with a link to the original CM code, in the "MATH_3_X" branch) When your work is merged in "master", we should not forget to add a link to the code in the [Geometry] repository. Thanks a lot for reporting the problem, Gilles >> [...] --------------------------------------------------------------------- To unsubscribe, e-mail: dev-unsubscr...@commons.apache.org For additional commands, e-mail: dev-h...@commons.apache.org
