Re: [Maxima-bugs] [ maxima-Bugs-1281737 ] limit(atan(x)/(1/exp(1)-exp(-(1+x)^2)),x,inf,plus) - wrong



I think it should be -i*pi, and so does Mathematica.
Commercial Macsyma thinks i*pi.
Numerical evaluation as x->0 from real positive values confirms
-i*pi is consistent with Maxima's numerics.
RJF


Raymond Toy wrote:

>>>>>>"gosei" == go furuya  writes:
>>>>>>            
>>>>>>
>
>    gosei> Hi Raimond
>    >> I've looked a bit more at this.  A simpler example is 
>    >> 
>    >> limit(log(-1-%i*x),x,0,plus);
>    >> 
>    >> Maxima returns log(-1) = %i*%pi.  However, the correct answer is
>    >> -%i*%pi because we're always in the third quadrant.
>
>    gosei> You may confused calculating limit in 2D with 1D.
>    gosei> In case of uniformization,both %i*%pi and -%i*%pi shoud be correct. 
>    gosei> All that is required is to select domain of uniformization
>    gosei> ,such that principal value.
>
>What does uniformization mean?  And why are they both correct?  What
>do you mean limit in 2D vs 1D?
>
>Perhaps I'm wrong in assuming log means principal log so the branch
>cut is the negative real axis.  
>
>Anyway, my reasoning is this:
>
>log(-1-%i*x) = 1/2*log(x^2+1) + %i*arg(-1-%i*x)
>
>As x -> 0 from above,  the realpart has limit 0.  For the imaginary part
>arg(-1-%i*x) is -pi + eps for small x > 0, so the limit is -pi.
>
>Thus the correct answer should be -%i*%pi.
>
>Ray
>
>_______________________________________________
>Maxima mailing list
>Maxima@math.utexas.edu
>http://www.math.utexas.edu/mailman/listinfo/maxima
>  
>