Numerical routines ready



On Fri, 2002-04-26 at 12:10, Raymond Toy wrote:
> 
> I think everything is in place now for the numerical routines from
> SLATEC.  I've tested this mostly with CMUCL and some with Clisp.  I
> can't test with gcl, so I hope some can do that for me.

Unfortunately, the GCL build is now broken because gcl does not have
defpackage. I think the solution to this is to steal a "defpackage" from
another lisp. If nobody else wants to do this, I'll give it a shot a
little later. I'm still finishing up my work on xmaxima.

> Notes:
> o You might have to run config.status.  Sometimes the dependencies
>   aren't built right or something and the files don't get compiled.
>   (I did not update clisp-depends.mk and cmucl-dpends.mk, assuming
>   that configure and/or make does that.)

The *-depends.mk should get checked in, but I can do that. There is
actually a reason to have them in cvs, but it's not worth getting
into...

> o The error reporting routines used by SLATEC are included and do
>   appear to work.  (Try gn(-20,60) for an example.)  May want to
>   change the routines to be more lisp-like and maybe dump the user
>   into the debugger for those fatal errors which now just return back
>   to the top-level loop after printing an error message.
> o The function in(x,n) always just returns unevaluated.  I don't know
>   how to fix that.  This doesn't happen in maxima 5.6.
> 
> What's next:
> o Some more work needs to be done on computing the values via simple
>   analytic continuation.
> o Maybe add the Bessel Y functions since the code is there.  Maybe K
>   and Hankel functions too?
> o More comments in the code.
> o Add documentation to the manual on what these functions are.
> o Add the new proposed names bessel_j, bessel_i, bessel_y, etc.?
> o Maybe tell maxima about some of the properties of the Bessel
>   functions, like derivatives?
> o Add access to the erfc function from maxima.  Right now you can only
>   numerically evaluate erf, but erfc is usually more important to me.
> o Replace the Gaussian rng with a better version from clocc.
> 
> Hope this is useful to someone,

I'm sure it will be useful to many. Thanks again.

--Jim