From: Franck Stauffer (franck_at_[hidden])
Date: 2006-12-04 13:14:27
Hello,
I have been using boost for quite a while, and I was wondering if
there is any interest (or ongoing development) in a numerical
quadrature library. I have seen some old discussions on the mailing
lists archive, but couldn't tell wether something was still in the
works.
I have coded the embryo of such a template library (for my personal
use) which basically works taking a unary_function and allows the
user (me in that case) to choose between different integration
algorithms (composite trapezoidal, simpson and romberg at the
moment). Performance is good (faster than the kind of implementation
you get from Numerical Recipes in C thanks to good inlining).
I also extended this to work on discrete data sets (such as a
vector<double> obtained from measurements etc...) rather than just
function with analytical expressions. 2D integration of
binary_function's can the previous 1D algorithms using functional's
binder1st, but I haven't done much in that direction to potentially
improve things.
I implemented two backbone classes to do the work but I am still
wondering what the "consensus" calling convention on that kind of
library could be (the choice of using a class was simply to be able
to go back on a previously integrated function and ask for a bit more
precision without having to compute points twice).
Any comments welcome,
All the best,
Franck