Bug report ID: 719832 - limit(exp(x*%i)*x, x, inf) should give infinity

Le 06/04/2010 23:40, Dieter Kaiser a ?crit :
> ...
>
> With this extension we will get the desired result for the example of
> the bug report:
>
> (%i1) limit(exp(%i*x)*x,x,inf);
> (%o1) infinity
>    
ok
> We get the same result for the following:
>
> (%i2) limit(exp(-%i*x)*x,x,inf);
> (%o2) infinity
>    
ok
> And as a consequence:
>
> (%i3) limit(sin(x)*x,x,inf);
> (%o3) infinity
>    
not ok : sin(x) vanishes infinitely often near inf
> (%i4) limit(cos(x)*x,x,inf);
> (%o4) infinity
>    
not ok !
> Two integrals in rtestint.mac will get a problem:
>
> ********************** Problem 59 ***************
> Input:
> integrate(cos(9*x^(7/3)),x,0,inf)
>
>
> Result:
> Principal Value
> 0
>
> This differed from the expected result:
> 3*gamma(3/7)*cos(3*%pi/14)/(7*9^(3/7))
>
> ********************** Problem 60 ***************
> Input:
> integrate(sin(9*x^(7/3)),x,0,inf)
>
>
> Result:
> 'integrate(sin(9*x^(7/3)),x,0,inf)
>
>
> The reason is, that Maxima no longer will get the limit '$und but
> '$infinity in the routine limit-pole for the involved trig expressions,
> e.g.
>
> (%i2) limit(x*cos(9*x^2),x,inf);
> (%o2) infinity
>
> But the routine limit-pole does not handle the case '$infinity and
> returns the default answer '$yes and not the expected result '$und. This
> could be corrected in the routine limit-pole by adding the case
> '$infinity with a result '$no:
>
> (defun limit-pole (exp var limit direction)
>    (let ((ans (cond ((member limit '($minf $inf) :test #'eq)
>                      (cond ((eq (special-convergent-formp exp limit)
> '$yes)
>                             '$no)
>                            (t (get-limit (m* exp var) var limit
> direction))))
>                     (t '$no))))
>      (cond ((eq ans '$no)   '$no)
>            ((null ans)   nil)
>            ((eq ans '$und) '$no)
> --->       ((eq ans '$infinity) '$no)
>            ((equal ans 0.)   '$no)
>            (t '$yes))))
>
> I am not sure about the following two points:
>
> 1. Is it correct to return the result '$infinity for the limits of
>     products with the trig functions sin and cos, e.g. for
>     limit(x*cos(x),x,inf) ->  infinity?
>    
no, x*cos(x) oscillates infinitely often between almost -inf and inf for 
x near inf

Eric Reyssat
> 2. The routine limit-pole seems not to expect a result '$infinity.
>     Is it correct in general to return the answer '$no for this case?
>
> Dieter Kaiser
>
>
> _______________________________________________
> Maxima mailing list
> Maxima at math.utexas.edu
> http://www.math.utexas.edu/mailman/listinfo/maxima
>
>