oldk1331 wrote:
> 
> I think it's doable.
> 
> First, by default, "limit" is for real expression (there are
> accompanying "complexLimit"), so it knows "sqrt(a)" is positive:
> 
> (8) -> limit(sqrt(a)*x,x=%plusInfinity)
> 
>     (8)   + infinity
> 
> So in theory, it should also compute for "erf(sqrt(a)*x)", but:
> 
> (9) -> limit(erf(sqrt(a)*x),x=%plusInfinity)
> 
>     (9)  "failed"
> 
> This should be very simple to compute, but we still use Gruntz
> algorithm to compute and we fail at there.  This may indicate
> a bug in Gruntz algorithm, and raise another question: shall
> we add a short path to handle simple cases, like "specialLimit"
> in limitps.spad?

Gruntz seem to be close to computing this limit, so I do not
see special reason to add alternate code path.  More generaly,
various "short paths" or "special paths" may seem attractive
at first, but they are pain in longer time.
 
> Another somewhat related problem:
> 
> (1) -> limit(sqrt(a^2)*x,x=%plusInfinity)
> 
>     (1)  "failed"
> 
> This case fails because "mrv_normalize" will destroy the structure
> of "sqrt(a^2)",

Do not attach too much importance to 'sqrt' in integral that
started this discussion:

                 +---+
        erf(a x)\|%pi
   (4)  --------------
              2 a
                                                    Type: Expression(Integer)

is a valid indefinite integral and we should be generating it
instead of version with 'sqrt(a^2)'.  Note that for this
expression limit splits into two cases: positive and negative
and _both_ give the same limit.  So we would need version of
Gruntz that can split and combine cases.

And for treating 'sqrt(a^2)' as 'abs(x)' we need version of
'normalize' that splits cases.

> so maybe a "assume" system will still be needed.
> But such a system is invasive and not always work in Maxima.

Yes, in longer term we will need "assume" and handling of
conditions.

-- 
                              Waldek Hebisch

-- 
You received this message because you are subscribed to the Google Groups 
"FriCAS - computer algebra system" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to fricas-devel+unsubscr...@googlegroups.com.
To post to this group, send email to fricas-devel@googlegroups.com.
Visit this group at https://groups.google.com/group/fricas-devel.
For more options, visit https://groups.google.com/d/optout.

Reply via email to