On Jul 16, 2012, at 4:36 AM, Sergiu Ivanov <[email protected]> wrote:
> Hello, > > On Mon, Jul 16, 2012 at 1:19 PM, [email protected] > <[email protected]> wrote: >> With all the issues surrounding ordering, hashes and hash >> randomization I was left with the impression that we should rewrite >> tests in such a way that they test for mathematical correctness but >> not for specific ordering of args, etc. > > Since we are discussing this, I would like to state that not forcing > the ordering in some cases leads to an unpleasant variety of results. > For example, in computing eigenvectors, this leads to _different_ > eigenvectors at different runs. They are not essentially different, > of course, but they are still different and this sounds a lot like a > concealed bomb. I'm curious, do these eigenvectors differ only by constants, or is it a case of completely different linear combinations with some eigenvalues of high geometric degree? > > I think that, on some occasions, certain containers absolutely must be > sorted to avoid such kind of instability. I think if we leave algorithms as such then they must be treated much in the same way as randomized algorithms (albeit randomized algorithms that don't have pseudo-false positives or negatives). Whether or not that's something that can or should be reasonably dealt with I'm not yet sure. Aaron Meurer > > Sergiu > > -- > You received this message because you are subscribed to the Google Groups > "sympy" group. > To post to this group, send email to [email protected]. > To unsubscribe from this group, send email to > [email protected]. > For more options, visit this group at > http://groups.google.com/group/sympy?hl=en. > -- You received this message because you are subscribed to the Google Groups "sympy" group. To post to this group, send email to [email protected]. To unsubscribe from this group, send email to [email protected]. For more options, visit this group at http://groups.google.com/group/sympy?hl=en.
