Arithmetic with CREs (including taylor series) is contagious. That is,
rat(x) + 1 == rat(x+1). It is even contagious for some function
applications, e.g. sin(taylor(x,x,0,10)) == taylor(sin(x),x,0,10). For
simple aggregates, it is not contagious: [x, rat(x)] preserves the first x
in general representation and the second in CRE representation.
For functions which are 'unaware' of CREs, the CRE is converted to general
representation, e.g.
f(rat(x)) => f(x)
If that were not done, then (e.g.) f(rat(x)) - f(x) would not simplify to 0.
-s
On Fri, Apr 29, 2011 at 14:27, Barton Willis <willisb at unk.edu> wrote:
> maxima-bounces at math.utexas.edu wrote on 04/29/2011 10:35:18 AM:
>
> > The taylor-ness of the expression was apparently thrown away
> > when it was the argument of some other operator.
>
> The same happens for CRE arguments:
>
> (%i8) foo(rat(x));
> (%o8) foo(x)
>
> (%i9) ?print(%);
> (($FOO SIMP) $X)
> (%o9) foo(x)
>
> I don't know what hunk of code does this.
>
> Simplifying functions that use the services of simpcheck instead of
> simplifya
> also squash CRE and taylor polynomial arguments.
>
> Some functions (trig simplification functions, for example), call
> taylorize to handle taylor arguments. I'd guess that the testsuite
> calls taylorize many many times--I hope it's fast---if it's not,
> blame me :(
>
> --Barton
> _______________________________________________
> Maxima mailing list
> Maxima at math.utexas.edu
> http://www.math.utexas.edu/mailman/listinfo/maxima
>