evaluating erf(%i) in maxima



Raymond Toy writes:
>>>>>>"Stavros" == Stavros Macrakis  writes:
> 
> 
>     Stavros> Hmm.  I guess I was too terse.  If you want to be completely consistent
>     Stavros> with base Maxima, it should not give a numerical value, since
>     Stavros> sin(0.5*%i+1.2),numer doesn't, either.  But that can be argued to be
>     Stavros> wrong.  So it seems fine to me that gamma(1+1/2*%i),numer should give a
>     Stavros> complex float, and maybe sin should be changed, too....
> 
> Oh, right.  I see what you're getting at.  And yes, it does annoy me
> that sin(1.2+0.5*%i) doesn't give a number, and it would be nice if
> that were changed.
> 
> Ray
> 

I think that it is quite important to make numerical
evaluation for complex and real numbers consistent.
And IMHO gamma in current Maxima CVS demonstrates
desired behavior:

gamma(real or complex inexact number) -> inexact number

gamma(real or complex exact number) -> gamma(...)
   or
gamma(real or complex exact number) -> exact number
   for special cases

gamma(real or complex exact number),numer -> inexact number

The only logical way to make all other elementary
functions sin, cos, ..., log behave in the same fashion.

BTW gamma(4),numer produces a bit unpleasant result
  0.0 %i + 6.0000000003

Strategical goal - complex numerical evaluation for
all elementary and special functions known to Maxima.
I know, this is hard and require a lot of work.



-- 
      Vadim V. Zhytnikov