This may be a snippet from a question I submitted.  Thanks for replying. There are many values of arguments for the Hypergeometric function for which it does not reduce to another special function. So in a nutshell that was the use case, a generic hypergeometric function that could not be expressed as another special function. Also, in my post I mentioned that the boost documentation listed these as unsupported TR1 functions.  Can you clarify what this means?  Does TR1 handle hypergeometric functions?  Thanks, David ________________________________ From: John Maddock To: boost-users@lists.boost.org Sent: Monday, March 19, 2012 4:38 AM Subject: Re: [Boost-users] Bessel and Hypergeometric functions wanted, was: (no subject)

...

2.  Does Boost plan to have hypergeometric functions in the future?

I've only found hypergeometric distributions: http://www.boost.org/doc/libs/1_48_0/libs/math/doc/sf_and_dist/html/math_too...

Right, I've deliberately avoided hypergeometetic functions, because, although they're theoretically attractive - you can represent anything as a hypergeometric after all - they're pretty much untestable (domain of functions with >3 args is too great for any kind of coverage), and except for the trivial use cases, the "obvious" implementations either don't converge or aren't numerically stable.  In other words it's better to stick with more "special" special functions with a more limited domain that can actually be tested/implemented in a reasonably comprehensive manner. What was the use case for the hypergeometrics BTW - was it to implement a special function that we don't otherwise have? HTH, John. _______________________________________________ Boost-users mailing list Boost-users@lists.boost.org http://lists.boost.org/mailman/listinfo.cgi/boost-users

Thanks again. I am familiar with gsl. I guess I will be writing some custom functions for the specific application I need. On another note I didn't see a post to question 1, does the bessel function family take complex argument? David Sent from my Verizon Wireless BlackBerry -----Original Message----- From: John Maddock Sender: boost-users-bounces@lists.boost.org Date: Mon, 19 Mar 2012 13:15:27 To: Reply-To: boost-users@lists.boost.org Subject: Re: [Boost-users] Bessel and Hypergeometric functions wanted, was: (no subject)

This may be a snippet from a question I submitted. Thanks for replying. There are many values of arguments for the Hypergeometric function for which it does not reduce to another special function. So in a nutshell that was the use case, a generic hypergeometric function that could not be expressed as another special function.

Also, in my post I mentioned that the boost documentation listed these as unsupported TR1 functions. Can you clarify what this means? Does TR1 handle hypergeometric functions?

Yes, they were added to the TR1 spec. However, that's part of the TR that did not make it into C++11 and to my knowledge no one has actually implemented them for the reasons I gave. I believe GSL has them (http://www.gnu.org/software/gsl/manual/html_node/Hypergeometric-Functions.ht...), but once you dig into the code there are lots of comments to the effect that they don't always behave well in some regions - and that's just the bits they've managed to test - to repeat, a function with > 3 real valued args is untestable over the whole domain, that's *why* they're broken down into more precisely defined functions. Regards, John. _______________________________________________ Boost-users mailing list Boost-users@lists.boost.org http://lists.boost.org/mailman/listinfo.cgi/boost-users