On Wed, 27 Feb 2013 15:19:34 Timo Kluck wrote: > Op woensdag 27 februari 2013 20:19:32 UTC+1 schreef Simon King het volgende: > > Hi Nils, > > > > On 2013-02-27, Nils Bruin <nbr...@sfu.ca <javascript:>> wrote: > > > Do we really want to test our random number generators for > > > deterministic behaviour? If so, the test should probably at least set > > > the seed prior to testing things that depend on actual values > > > produced. > > > > Isn't this the case already? > > In this case, it is. The random number seed is even explicitly set (to > 1337) in the doctest itself. > > > > Of course, if the algorithm used for the pseudo-random > > > generator has actually changed, setting the seed would not be enough, > > > but that should be rare. > > > > If I understand correctly, that rare situation occurs here. Hence, one can > > still rely on deterministic output, but a *different* deterministic > > output. > > Exactly. > > > > The more appropriate thing might be to rewrite the test to check > > > general behaviour rather than rely on exact output. > > > > +1 > > > > Testing against theorems (such as: One has random data, but the test is > > using an identity that must hold for any data) is most elegant. > > > > I agree. I don't know much about polybori though, so I can't really change > > those doctests in that way. > > My question is mostly: I assess the situation as being such as you describe > above: there is a rare situation where we should accept finding different, > yet deterministic, output. However, maybe a polybori expert might correct > me and say that this random_set function is often used in such a way that > it is vital that it stays backward compatible. I have a hard time imagining > that, but I'd better be safe than sorry. > > It would be great if someone who actually knows what anyone would want to > use random_set for, to give us the go-ahead. >
Actually because in sage-on-gentoo we use the system boost we hit that particular doctest failure a long time ago. I then asked Alexander Dreyer who works on polybori if the output was ok and it didn't seem to be concerned. You can give him a ping for his opinion. Martin Albretch could also be contacted. Francois -- You received this message because you are subscribed to the Google Groups "sage-devel" group. To unsubscribe from this group and stop receiving emails from it, send an email to sage-devel+unsubscr...@googlegroups.com. To post to this group, send email to sage-devel@googlegroups.com. Visit this group at http://groups.google.com/group/sage-devel?hl=en. For more options, visit https://groups.google.com/groups/opt_out.
