[Date Index] [Thread Index] [Author Index]
Re: Possible bug in Floor function?
On Sat, Feb 22, 2014 at 03:03:53AM -0500, John Doty wrote: > On Wednesday, February 19, 2014 1:43:20 AM UTC-5, Itai Seggev wrote: > > > It's not a bug per se, but it is a somewhat unfortunate edge case. No one has > > > > really hit on the fundamental issue, > > I hit on it in my post in this thread, and you didn't address one one the Yes you did, although your email made it to the list after I had sent my email. > important details here. Let me repeat: > > Floor seems to have no trouble with other hairy expressions that happen to equal integers, for example "((-1)^(1/3) + (-1)^(2/3))/Sqrt[-3]". But a mere Simplify reduces that to 1. Perhaps that's where the real bug is. Does Floor try Simplify? Maybe Simplify should be transforming things like Log/Log. No, Floor doesn't use Simplify. It uses adaptive numerical approximation. It's just that for your hidden integer, it gets an answer it can verify (on its first attempt, no less), whereas for the Log case that started the threade it can't verify its answer and defaults to raw machine arithmetic. As to whether Simplify should undo Logs, I don't know. I don't know enough about the internals for Simplify vs FullSimplify and where we draw the line to comment intelligently on that point. I can certainly pass it along as suggestion to the relevant developer. -- Itai Seggev Mathematica Algorithms R&D 217-398-0700