Re: [Reduce-algebra-developers] Bug in sum?

[Rainer S. <rai...@gm...>]

On Thu, 23 Aug 2012 at 20:24 +0100, Arthur Norman wrote:

 > Thank you for both the bug report and for pointing out exactly where 
 > something needed changing. The problem arose because a floating point 
 > number ends up represented internally by something that is a "sort of" 
 > prefix form but the data in it is chain terminating on a number not on 
 > "nil". So freeof!-dfl [as you observed] crashed on it.

Umm. This gets around the crashing, but freeof!-df doesn't process 
correctly special prefix forms in its first parameter.

Consider:

load_package sum;
operator f;

lisp freeof!-df(aeval '(df (f u v) u v),'v); % -> t
lisp freeof!-df(reval '(df (f u v) u v),'v); % -> nil

although both should evaluate to nil.

So I propose something like

 symbolic procedure freeof!-df(u, v);
   % check u contains differential operator with respect to v;
   if atom u or car u eq '!:dn!: or get(car u,'dname) then t
    else if car u eq '!*sq then freeof!-df(prepsq cadr u,v)
    else if car(u) eq 'df
     then freeof!-df(cadr u,v) and not smember(v,cddr u)
    else freeof!-dfl(cdr u,v);


  Rainer