On 11/22/2012 03:57 PM, Raymond Toy wrote:
>>>>>> "Jaime" == Jaime Villate <villate at fe.up.pt> writes:
> Jaime> On 11/21/2012 11:58 PM, Raymond Toy wrote:
> Stavros> On Wed, Nov 21, 2012 at 4:14 PM, Edwin Woollett <woollett at charter.net> wrote:
> >>
> >>
> Stavros> True, Maxima stupidly fails on integrate(1/(1+sqrt(x)),x,0,1),
> Stavros> but integrate(ratsimp(1/(1+sqrt(x))),x,0,1), algebraic:true does work....
> >>
> >> Yes, as mentioned in a different thread, the cause is a failure in
> >> poles-in-interval to determine that there are no poles in the
> >> interval. Setting intanalysis:false allows maxima to correctly return
> >> 2-2*log(2);
> >>
> Jaime> It would be nice that before returning a noun form integrate() tried
> Jaime> setting intanalysis:false or algebraic:true.
>
> I don't think that should be done in general, but it might make sense
> to do that when poles-in-interval returns unknown, which is what
> happens in this case. However, if there actually were a pole in the
> interval, then result is very likely to be wrong because maxima will
> just substitute in the limits of integration into the integrand,
> ignoring any singularities. That would probably be worse than
> returning a noun form.
>
I was thinking just in cases like this one, 1/(1+sqrt(x))). That
function do not really have poles in the interval [x,0,1]. It is its
derivative the one which has a pole.
Jaime