Hi Yannick, > > About the default value (1e-14), I know the tolerance is case-dependant, but > some times, we don't precisely know the order of magnitude (ex : jacobians > matrices) and it's quite practical to have a default value. > In our tools, we use this value in the test classes, but also for specific > algorithm convergence threshold. > I'm uncomfortable with CM having a default tolerance implemented somewhere. As you said, tolerances are case-dependant, and therefore DEFAULT_TOL (or whatever you want to call it) should stand in your particular application, not in a utility Commons-Math class. This is pretty much the approach adopted in our testing suites. You will frequently find a constant field with a default tolerance which applies for the specific unit test only. I would be very reluctant to try and be more general, while we all know that such a constant would not mean anything.
Don't mistake me. I'm not saying that chosing a (somewhat arbitrary) default tolerance is bad practice. As you said, sometimes there is no other option. I would like to know what others think, Sébastien --------------------------------------------------------------------- To unsubscribe, e-mail: dev-unsubscr...@commons.apache.org For additional commands, e-mail: dev-h...@commons.apache.org
