Re: [julia-users] hankelh1(1,z) bug?

2016-05-26 Thread Isaiah Norton
This looks like an issue in the underlying AMOS library from netlib. Both
SciPy and SciLab give the same answer as Julia and also use AMOS. mpmath
and Wolfram Alpha give the correct answer. Possibly a documentation issue
about tolerances in the AMOS implementation, I'm not quite sure.

Issue: https://github.com/JuliaLang/openspecfun/issues/45

On Wed, May 25, 2016 at 2:51 PM, Edmundo Lavia <
sivasadartantasvueltasnosi...@gmail.com> wrote:

> It seems to me that the hankelh1 function ([image: H_\nu(z)]) has a bug
> for small argument values,
> for example:
>
> > hankelh1(1,2e-32)
> 1.9490859162596872e15 - 3.1830988618379058e31im
>
> while
>
> > besselj(1,2e-32) + im*bessely(1,2e-32)
> 1.0e-32 - 3.1830988618379067e31im
>
> Iam using Version 0.5.0-dev+3667 under Fedora Linux 23.
> Cheers,
> E.
>


[julia-users] hankelh1(1,z) bug?

2016-05-25 Thread Edmundo Lavia
It seems to me that the hankelh1 function ([image: H_\nu(z)]) has a bug for
small argument values,
for example:

> hankelh1(1,2e-32)
1.9490859162596872e15 - 3.1830988618379058e31im

while

> besselj(1,2e-32) + im*bessely(1,2e-32)
1.0e-32 - 3.1830988618379067e31im

Iam using Version 0.5.0-dev+3667 under Fedora Linux 23.
Cheers,
E.