[SciPy-dev] scipy.special.hyp2f1 bug?

[Fredrik Johansson]

On Thu, Sep 17, 2009 at 8:54 PM, Fredrik Johansson
<fredrik.johansson at gmail.com> wrote:
> Which convention makes more sense? I'd be happy to change mpmath
> (though I need to check how many derived results are changed). It
> would probably be easy to support both conventions in mpmath by adding
> a flag.

I checked, and the only tests in mpmath that break if this change is
made are explicit tests for the value of hyp2f1(a,-n,-n,z). I'm
tempted to simply change mpmath's behavior and add a documentation
note to the effect that Mathematica is wrong :-)

Fredrik