[SciPy-dev] Issues in sici

David Cournapeau cournape@gmail....
Fri Nov 27 03:02:42 CST 2009


On Fri, Nov 27, 2009 at 5:38 AM, Arkapravo Bhaumik
<arkapravobhaumik@gmail.com> wrote:
> Hi everyone
>
> I have come across a minor issue with the sici module which evaluates the
> sine integral. Evaluation is excellent and gives precise values, however for
> inputs of infinity the results are nan which is very annoying.

Yes, that's a bug in cephes: the case for big x uses cos and sin,
which is not a good in the case of infinite input. I added a special
case to return the theoretical values in the case of real input,

See #1060 for details, fixed in r6122. Note that the function is still
wrong for negative values, the real fix is more likely to rewrite the
function, as cephes quality for special case is generally quite poor
(and the code horrible).

cheers,

David


More information about the SciPy-Dev mailing list