On Wed, Apr 22, 2009 at 6:54 PM, Carl Witty <carl.wi...@gmail.com> wrote: > > > Would it be better to test the results numerically? (For instance, > evaluate the integral returned and the desired result at 100 random > points to high precision, and ensure that the relative error between > the answers at each point is small.)
How can one do that with symbolic variables? Most of the test integrals have symbolic constants (w.r.t the integration) so it isn't just the integration variable. I thought about numerical testing, but it isn't generally feasible. > > Of course, this wouldn't count as a proof that the result was correct, > but IMHO it would be good enough (it seems unlikely that integration > bugs would result in wrong answers that were numerically almost > equivalent to the right answer). (Actually, I might actually trust a > numeric result more than a symbolic simplification-based result, given > the theoretical possibility that a simplification bug might cancel out > an integration bug, leading to a false pass in the test suite; > especially if simplification and integration are done in the same > system.) It's possible to have simplification bugs, but I'll have to rely upon separate tests of the simplification system. Cheers, Tim Lahey. --~--~---------~--~----~------------~-------~--~----~ To post to this group, send email to sage-devel@googlegroups.com To unsubscribe from this group, send email to sage-devel-unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/sage-devel URLs: http://www.sagemath.org -~----------~----~----~----~------~----~------~--~---
