Re: [Boost-users] [mpl::math] exponential metafunction for complex types

From: Cromwell Enage (sponage_at_[hidden])
Date: 2007-02-21 15:08:43

--- Chris Weed wrote:
> Hi,
> I am getting the wrong result for the exponential of
> complex numbers such as the following:
> 
> #include <boost/mpl/math/double.hpp>
> #include <boost/mpl/math/complex_number.hpp>
> #include <iostream>
> 
> int main()
> {
>   typedef
>
boost::mpl::math::string_c_to_double<'-',2,'.',0>::type
> minus_two;
>   typedef
>
boost::mpl::times<minus_two,boost::mpl::math::pi<boost::mpl::math::double_tag>
> >::type minus_two_pi;
>   typedef
>
boost::mpl::math::complex_number<double_zero,minus_two_pi>
> imag_minus_two_pi;
>   std::cout << imag_minus_two_pi() << std::endl;
> 
>   std::cout <<
>
boost::mpl::math::exponential<imag_minus_two_pi>::type()
> << std::endl;
>   return 0;
> }
> 
> This prints:
> 0-6.28319i
> 1.0329-0.0109834i
> 
> which should be
> 0-6.28319i
> 1.0000-0.00000i
> 
> Chris

The problem was in the sine and cosine functions,
which by default lose their accuracy when the angle is
beyond tha range (-pi/2,pi/2).  I've fixed their
implementations to take advantage of the fact that
they are periodic functions; however, this also
reintroduces recursive template instantiation.

Your program now resides in the example directory as
complex_exponential.cpp; note that I've changed the
semantics of the tagged constants.  The metafunctions
pi<> et al. take in primitive numeric types, e.g. int,
double; if you wish to pass in numeric tags, use
pi_dispatch<>, etc.

                              Cromwell D. Enage

 
____________________________________________________________________________________
Do you Yahoo!?
Everyone is raving about the all-new Yahoo! Mail beta.
http://new.mail.yahoo.com