On 10/28/11 12:17 PM, Edwin Woollett wrote:
> Despite using the new control features to surpress
> slatec error messages, here is a quad_qagi example
> which appears to generate a error message.
>
> (I am using load ("j4save.lisp")$
> and quad_j4save('control, 0)$ in nint.mac until
> the next Windows binary version of Maxima comes
> out.)
> -------------
> (%i1) load(nint);
> (%o1) "c:/work2/nint.mac"
>
> (%i8)
> quad_qagi((sin(x+sqrt(x))+x*bessel_j(0,x^2))/(1+x),x,0,inf,limit=700);
>
> * NO PRECISION BECAUSE X IS BIG
>From looking at the code, this looks like it's intentional. To fix this
the right way, the slatec Fortran code would need to be changed.
>
> Any good ideas for using slatec (or other) to get the finite answer?
The quadpack manual had some examples of integrating bessel functions.
Basically, integrate between zeroes of the bessel functions, and sum the
results, possibly using some kind of convergence acceleration technique
for the sum. Not sure how effective this would for this, since it's not
just f(x)*bessel(n,x).
Ray